SERVER-109844 Basic support for disaggregated storage clusters (#40566)

Co-authored-by: Benety Goh <benety@mongodb.com>
Co-authored-by: Mathias Stearn <mathias@mongodb.com>
Co-authored-by: Kaitlin Mahar <kaitlin.mahar@mongodb.com>
Co-authored-by: Brandon Stoll <bstoll@users.noreply.github.com>
Co-authored-by: Vanessa Noia <54818020+nessnoia@users.noreply.github.com>
Co-authored-by: graphite-app[bot] <96075541+graphite-app[bot]@users.noreply.github.com>
Co-authored-by: Vishnu K <vishnu.kaushik@mongodb.com>
Co-authored-by: Sunil Narasimhamurthy <suniltheta@gmail.com>
Co-authored-by: Jiawei Yang <youngyang0820@gmail.com>
Co-authored-by: Will Korteland <korteland@users.noreply.github.com>
Co-authored-by: Saman Memaripour <amirsaman.memaripour@mongodb.com>
Co-authored-by: huayu-ouyang <huayu.ouyang@mongodb.com>
Co-authored-by: Suganthi Mani <38441312+smani87@users.noreply.github.com>
Co-authored-by: Thomas Goyne <thomas.goyne@mongodb.com>
Co-authored-by: Haley Connelly <haley.connelly@mongodb.com>
Co-authored-by: Billy Donahue <BillyDonahue@users.noreply.github.com>
Co-authored-by: Kirollos Morkos <kiro.morkos@mongodb.com>
Co-authored-by: Lingzhi Deng <lingzhi.deng@mongodb.com>
Co-authored-by: Hartek Sabharwal <hartek.sabharwal@mongodb.com>
Co-authored-by: Aaron Himelman <aaron.himelman@mongodb.com>
Co-authored-by: Moustafa Maher <m.maher@mongodb.com>
Co-authored-by: prathmesh-kallurkar <prathmesh.kallurkar@mongodb.com>
Co-authored-by: Dan Larkin-York <13419935+dhly-etc@users.noreply.github.com>
Co-authored-by: Shreyas Kalyan <35750327+shreyaskalyan@users.noreply.github.com>
Co-authored-by: Shreyas Kalyan <shreyas.kalyan@mongodb.com>
Co-authored-by: Jonathan Reams <jbreams@mongodb.com>
Co-authored-by: adriangzz <adriangonzalezmontemayor@gmail.com>
Co-authored-by: Eric Milkie <milkie@users.noreply.github.com>
Co-authored-by: Aaron B <aaron.balsara@mongodb.com>
Co-authored-by: Ali Mir <ali.mir@mongodb.com>
Co-authored-by: Alex Blekhman <alexander.blekhman@mongodb.com>
Co-authored-by: mpobrien <mpobrien005@gmail.com>
Co-authored-by: Mark Benvenuto <mark.benvenuto@mongodb.com>
Co-authored-by: Ruby Chen <ruby.chen@mongodb.com>
Co-authored-by: Jagadish Nallapaneni <146780625+jagadishmdb@users.noreply.github.com>
Co-authored-by: Jonas Bergler <jonas.bergler@mongodb.com>
Co-authored-by: Peter Macko <peter.macko@mongodb.com>
Co-authored-by: Nic <nic.hollingum@mongodb.com>
Co-authored-by: Jiawei Yang <jiawei.yang@mongodb.com>
Co-authored-by: Jordi Serra Torrens <jordist@users.noreply.github.com>
Co-authored-by: Sunil Narasimhamurthy <sunil.narasimhamurthy@mongodb.com>
GitOrigin-RevId: a1c6609c820052137e2aa759711e86c337ae6f9f

.github/CODEOWNERS

@@ -81,6 +81,9 @@ WORKSPACE.bazel @10gen/devprod-build @svc-auto-approve-bot
 /buildscripts/idl/**/idl_compatibility_errors.py @10gen/query-optimization @svc-auto-approve-bot
 /buildscripts/idl/**/test_compatibility.py @10gen/query-optimization @svc-auto-approve-bot
 
+# The following patterns are parsed from ./buildscripts/modules/atlas/OWNERS.yml
+/buildscripts/modules/atlas/ @10gen/server-disagg-storage @svc-auto-approve-bot
+
 # The following patterns are parsed from ./buildscripts/monitor_build_status/OWNERS.yml
 /buildscripts/monitor_build_status/ @10gen/devprod-correctness @svc-auto-approve-bot
 
@@ -888,6 +891,8 @@ WORKSPACE.bazel @10gen/devprod-build @svc-auto-approve-bot
 /jstests/libs/**/catalog_list_operations_consistency_validator.js @10gen/server-catalog-and-routing @svc-auto-approve-bot
 /jstests/libs/**/raw_operation_utils.js @10gen/server-collection-write-path @svc-auto-approve-bot
 /jstests/libs/**/json_utils.js @10gen/query-integration-extensions @svc-auto-approve-bot
+/jstests/libs/**/replicated_ident_utils.js @10gen/server-storage-engine-integration @svc-auto-approve-bot
+/jstests/libs/**/replicated_record_ids_utils.js @10gen/server-storage-engine-integration @svc-auto-approve-bot
 
 # The following patterns are parsed from ./jstests/libs/clustered_collections/OWNERS.yml
 /jstests/libs/clustered_collections/**/* @10gen/server-collection-write-path @svc-auto-approve-bot
@@ -1831,9 +1836,6 @@ WORKSPACE.bazel @10gen/devprod-build @svc-auto-approve-bot
 /src/mongo/db/commands/query_cmd/**/release_memory_cmd.* @10gen/query-execution @svc-auto-approve-bot
 /src/mongo/db/commands/query_cmd/**/update_metrics.* @10gen/query-execution @svc-auto-approve-bot
 
-# The following patterns are parsed from ./src/mongo/db/disagg_storage/OWNERS.yml
-/src/mongo/db/disagg_storage/**/* @10gen/server-disagg-storage @svc-auto-approve-bot
-
 # The following patterns are parsed from ./src/mongo/db/exec/OWNERS.yml
 /src/mongo/db/exec/**/* @10gen/query-execution-classic @svc-auto-approve-bot
 /src/mongo/db/exec/**/OWNERS.yml @10gen/query-execution-staff-leads @10gen/query-integration-staff-leads @10gen/query-optimization-staff-leads @svc-auto-approve-bot
@@ -2043,6 +2045,15 @@ WORKSPACE.bazel @10gen/devprod-build @svc-auto-approve-bot
 # The following patterns are parsed from ./src/mongo/db/modules/atlas/OWNERS.yml
 /src/mongo/db/modules/atlas/**/* @10gen/server-disagg-storage @svc-auto-approve-bot
 
+# The following patterns are parsed from ./src/mongo/db/modules/atlas/jstests/disagg_storage/OWNERS.yml
+/src/mongo/db/modules/atlas/jstests/disagg_storage/**/* @10gen/server-disagg-storage @svc-auto-approve-bot
+
+# The following patterns are parsed from ./src/mongo/db/modules/atlas/src/disagg_storage/OWNERS.yml
+/src/mongo/db/modules/atlas/src/disagg_storage/**/* @10gen/server-disagg-storage @svc-auto-approve-bot
+
+# The following patterns are parsed from ./src/mongo/db/modules/atlas/src/disagg_storage/encryption/OWNERS.yml
+/src/mongo/db/modules/atlas/src/disagg_storage/encryption/**/* @10gen/server-security @svc-auto-approve-bot
+
 # The following patterns are parsed from ./src/mongo/db/modules/enterprise/OWNERS.yml
 /src/mongo/db/modules/enterprise/BUILD.bazel @10gen/devprod-build @svc-auto-approve-bot
 /src/mongo/db/modules/enterprise/README.md @10gen/server-release @svc-auto-approve-bot
@@ -2589,6 +2600,9 @@ WORKSPACE.bazel @10gen/devprod-build @svc-auto-approve-bot
 # The following patterns are parsed from ./src/mongo/db/repl/split_horizon/OWNERS.yml
 /src/mongo/db/repl/split_horizon/**/* @10gen/server-split-horizon @svc-auto-approve-bot
 
+# The following patterns are parsed from ./src/mongo/db/rss/OWNERS.yml
+/src/mongo/db/rss/**/* @10gen/server-replication @10gen/server-storage-execution @svc-auto-approve-bot
+
 # The following patterns are parsed from ./src/mongo/db/s/OWNERS.yml
 /src/mongo/db/s/**/* @10gen/server-cluster-scalability @svc-auto-approve-bot
 /src/mongo/db/s/**/*transaction* @10gen/server-transactions @svc-auto-approve-bot
@@ -3068,6 +3082,7 @@ WORKSPACE.bazel @10gen/devprod-build @svc-auto-approve-bot
 /src/third_party/**/croaring @10gen/query-execution @svc-auto-approve-bot
 /src/third_party/**/fmt @10gen/server-programmability @svc-auto-approve-bot
 /src/third_party/**/folly @10gen/server-workload-scheduling @svc-auto-approve-bot
+/src/third_party/**/googletest_restricted_for_disagg_only @10gen/server-disagg-storage @svc-auto-approve-bot
 /src/third_party/**/gperftools @10gen/server-workload-scheduling @svc-auto-approve-bot
 /src/third_party/**/grpc @10gen/server-networking-and-observability @svc-auto-approve-bot
 /src/third_party/**/icu4c* @10gen/query-execution @svc-auto-approve-bot

README.third_party.md

@@ -66,6 +66,7 @@ a notice will be included in
 | [pyiso8601]                                          | MIT                                                                                                 | 2.1.0                                    | unknown              |                                 |
 | [RoaringBitmap/CRoaring]                             | Unknown License                                                                                     | v3.0.1                                   |                      | ✗                               |
 | [SchemaStore/schemastore]                            | Apache-2.0                                                                                          | Unknown                                  |                      |                                 |
+| [sls-proto]                                          | Unknown License                                                                                     | 1.0                                      | unknown              | ✗                               |
 | [smhasher]                                           | Unknown License                                                                                     | Unknown                                  | unknown              | ✗                               |
 | [Snowball Stemming Algorithms]                       | BSD-3-Clause                                                                                        | 7b264ffa0f767c579d052fd8142558dc8264d795 | ✗                    | ✗                               |
 | [subunit]                                            | BSD-3-Clause, Apache-2.0                                                                            | 1.4.4                                    | unknown              |                                 |
@@ -122,6 +123,7 @@ a notice will be included in
 [opentelemetry-cpp]: https://github.com/open-telemetry/opentelemetry-cpp/
 [opentelemetry-proto]: https://github.com/open-telemetry/opentelemetry-proto
 [pyiso8601]: https://pypi.org/project/iso8601/
+[sls-proto]: https://github.com/10gen/sls
 [smhasher]: https://github.com/aappleby/smhasher/blob/a6bd3ce/
 [subunit]: https://github.com/testing-cabal/subunit
 [tcmalloc]: https://github.com/google/tcmalloc

buildscripts/resmokeconfig/resmoke_modules.yml

@@ -1,12 +1,12 @@
 enterprise:
   jstest_dirs:
     - src/mongo/db/modules/enterprise/jstests
-# atlas:
+atlas:
-#   fixture_dirs:
+  fixture_dirs:
-#     - buildscripts/modules/atlas/fixtures
+    - buildscripts/modules/atlas/fixtures
-#   hook_dirs:
+  hook_dirs:
-#     - buildscripts/modules/atlas/hooks
+    - buildscripts/modules/atlas/hooks
-#   suite_dirs:
+  suite_dirs:
-#     - buildscripts/modules/atlas/suites
+    - buildscripts/modules/atlas/suites
-#   jstest_dirs:
+  jstest_dirs:
-#     - buildscripts/modules/atlas/jstests
+    - src/mongo/db/modules/atlas/jstests

buildscripts/resmokelib/testing/fixtures/_builder.py

@@ -1,5 +1,6 @@
 """Utilities for constructing fixtures that may span multiple versions."""
 
+import json
 import logging
 import threading
 from abc import ABC, abstractmethod
@@ -223,6 +224,33 @@ class ReplSetBuilder(FixtureBuilder):
             )
             replset.install_mongod(node)
 
+        if replset.disagg_base_config:
+            members = []
+            for idx, node in enumerate(replset.nodes):
+                member = {
+                    "_id": idx,
+                    "host": node.get_internal_connection_string(),
+                    "priority": 1
+                }
+                members.append(member)
+            disagg_base_config = {
+                **replset.disagg_base_config,
+                "replSetConfig": {
+                    "_id": replset.replset_name,
+                    "version": 1,
+                    "term": 1,
+                    "members": members,
+                }
+            }
+            for node in replset.nodes:
+                opts = node.get_mongod_options()
+                opts["set_parameters"]["disaggregatedStorageConfig"] = json.dumps(
+                    disagg_base_config)
+                opts["set_parameters"]["disaggregatedStorageEnabled"] = True
+                opts["set_parameters"]["logComponentVerbosity"] = json.dumps(
+                    {"disaggregatedStorage": 5})
+                node.set_mongod_options(opts)
+
         if replset.start_initial_sync_node:
             if not replset.initial_sync_node:
                 replset.initial_sync_node_idx = replset.num_nodes

buildscripts/resmokelib/testing/fixtures/replicaset.py

@@ -74,6 +74,7 @@ class ReplicaSetFixture(interface.ReplFixture, interface._DockerComposeInterface
         launch_mongot=False,
         load_all_extensions=False,
         router_endpoint_for_mongot: Optional[int] = None,
+        disagg_base_config=None,
     ):
         """Initialize ReplicaSetFixture."""
 
@@ -139,6 +140,8 @@ class ReplicaSetFixture(interface.ReplFixture, interface._DockerComposeInterface
         # Set the default oplogSize to 511MB.
         self.mongod_options.setdefault("oplogSize", 511)
 
+        self.disagg_base_config = disagg_base_config
+
         # The dbpath in mongod_options is used as the dbpath prefix for replica set members and
         # takes precedence over other settings. The ShardedClusterFixture uses this parameter to
         # create replica sets and assign their dbpath structure explicitly.
@@ -462,12 +465,14 @@ class ReplicaSetFixture(interface.ReplFixture, interface._DockerComposeInterface
         primary = self.nodes[0]
         client = primary.mongo_client()
         while True:
-            self.logger.info("Waiting for primary on port %d to be elected.", primary.port)
+            self.logger.info(
-            is_master = client.admin.command("isMaster")["ismaster"]
+                "Waiting for primary on port %d to be elected.", primary.port)
-            if is_master:
+            cmd_result = client.admin.command("isMaster")
+            if cmd_result["ismaster"]:
                 break
             time.sleep(0.1)  # Wait a little bit before trying again.
-        self.logger.info("Primary on port %d successfully elected.", primary.port)
+        self.logger.info(
+            "Primary on port %d successfully elected.", primary.port)
 
     def _await_secondaries(self):
         # Wait for the secondaries to become available.

buildscripts/resmokelib/testing/fixtures/standalone.py

@@ -188,6 +188,12 @@ class MongoDFixture(interface.Fixture, interface._DockerComposeInterface):
             self.logger.debug("Mongod not running when gathering standalone fixture pid.")
         return out
 
+    def get_mongod_options(self):
+        return self.mongod_options
+
+    def set_mongod_options(self, options):
+        self.mongod_options = options
+
     def _handle_await_ready_retry(self, deadline):
         remaining = deadline - time.time()
         if remaining <= 0.0:

buildscripts/resmokelib/testing/testcases/fixture.py

@@ -50,6 +50,10 @@ class FixtureSetupTestCase(FixtureTestCase):
             self.fixture.await_ready()
             if (
                 not isinstance(self.fixture, (fixture_interface.NoOpFixture, ExternalFixture))
+                # TODO(SERVER-109851): Remove this.
+                # disagg mongod does not yet support "refreshLogicalSessionCacheNow" because it requires
+                # wtimeout support.
+                and self.fixture.__class__.__name__ != "DisaggFixture"
                 # Replica set with --configsvr cannot run refresh unless it is part of a sharded cluster.
                 and not (
                     isinstance(self.fixture, ReplicaSetFixture)

buildscripts/resmokelib/testing/testcases/jstest.py

@@ -3,7 +3,9 @@
 import copy
 import os
 import os.path
+import random
 import shutil
+import string
 import sys
 import threading
 import uuid
@@ -67,6 +69,9 @@ class _SingleJSTestCase(interface.ProcessTestCase):
         global_vars["MongoRunner.dataPath"] = data_path
 
         test_data = global_vars.get("TestData", {}).copy()
+        test_run_id = "".join(random.choices(string.ascii_letters + string.digits, k=10))
+        self.fixture.test_run_id = test_run_id
+        test_data["test_run_id"] = test_run_id
         test_data["minPort"] = core.network.PortAllocator.min_test_port(self.fixture.job_num)
         test_data["maxPort"] = core.network.PortAllocator.max_test_port(self.fixture.job_num)
         test_data["peerPids"] = self.fixture.pids()

buildscripts/tests/burn_in_tests_end2end/__init__.py

@@ -1 +0,0 @@
-"""Empty."""

buildscripts/tests/burn_in_tests_end2end/test_burn_in_tests_end2end.py

@@ -1,51 +0,0 @@
-import os
-import subprocess
-import sys
-import unittest
-
-import yaml
-
-import buildscripts.burn_in_tests as under_test
-
-
-class TestBurnInTestsEnd2End(unittest.TestCase):
-    @unittest.skip(
-        "Disabled since this test has behavior dependent on currently modified jstests. Re-enable with SERVER-108783."
-    )
-    @classmethod
-    def setUpClass(cls):
-        subprocess.run(
-            [
-                sys.executable,
-                "buildscripts/burn_in_tests.py",
-                "generate-test-membership-map-file-for-ci",
-            ]
-        )
-
-    @classmethod
-    def tearDownClass(cls):
-        if os.path.exists(under_test.BURN_IN_TEST_MEMBERSHIP_FILE):
-            os.remove(under_test.BURN_IN_TEST_MEMBERSHIP_FILE)
-
-    def test_valid_yaml_output(self):
-        process = subprocess.run(
-            [
-                sys.executable,
-                "buildscripts/burn_in_tests.py",
-                "run",
-                "--yaml",
-            ],
-            text=True,
-            capture_output=True,
-        )
-        self.assertEqual(
-            0,
-            process.returncode,
-            process.stderr,
-        )
-
-        output = process.stdout
-        try:
-            yaml.safe_load(output)
-        except Exception:
-            self.fail(msg="burn_in_tests.py does not output valid yaml.")

etc/evergreen.yml

@@ -93,8 +93,6 @@ include:
 
   - filename: etc/evergreen_yml_components/variants/codecoverage/test_dev.yml
 
-  - filename: src/mongo/db/modules/atlas/atlas_dev.yml
-
 parameters:
   - key: evergreen_config_file_path
     value: "etc/evergreen.yml"

etc/evergreen_lint.yml

@@ -79,6 +79,7 @@ rules:
           - assigned_to_jira_team_server_workload_scheduling
           - assigned_to_jira_team_server_networking_and_observability
           - assigned_to_jira_team_server_integration
+          - assigned_to_jira_team_server_disagg
 
           # https://github.com/10gen/mothra/blob/main/mothra/teams/rnd_dev_prod.yaml
           - assigned_to_jira_team_devprod_build
@@ -130,6 +131,7 @@ rules:
           - incompatible_inmemory
           - incompatible_all_feature_flags
           - incompatible_development_variant
+          - incompatible_disaggregated_storage
           - requires_compile_variant
           - requires_large_host
           - requires_large_host_tsan

etc/evergreen_yml_components/definitions.yml

@@ -1236,6 +1236,19 @@ functions:
       args:
         - "./src/evergreen/resmoke_tests_execute_bazel.sh"
 
+  "assume ECR role": &assume_ecr_role
+    command: ec2.assume_role
+    params:
+      role_arn: "${disagg_storage_ecr_arn}"
+
+  "fetch module images": &fetch_module_images
+    command: subprocess.exec
+    params:
+      binary: bash
+      add_expansions_to_env: true # needed to get the AWS secrets from ec2.assume_role
+      args:
+        - "./src/evergreen/fetch_module_images.sh"
+
   "retrieve generated test configuration":
     &retrieve_generated_test_configuration
     command: s3.get
@@ -1472,6 +1485,8 @@ functions:
     - *f_expansions_write
     - *sign_macos_dev_binaries
     - *multiversion_exclude_tags_generate
+    - *assume_ecr_role
+    - *fetch_module_images
     - *execute_resmoke_tests
       # The existence of the "run_tests_infrastructure_failure" file indicates this failure isn't
       # directly actionable. We use type=setup rather than type=system or type=test for this command
@@ -1521,6 +1536,8 @@ functions:
     - *configure_evergreen_api_credentials
     - *sign_macos_dev_binaries
     - *multiversion_exclude_tags_generate
+    - *assume_ecr_role
+    - *fetch_module_images
     - *execute_resmoke_tests
       # The existence of the "run_tests_infrastructure_failure" file indicates this failure isn't
       # directly actionable. We use type=setup rather than type=system or type=test for this command
@@ -3294,3 +3311,11 @@ functions:
           [AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN]
         args:
           - "./src/evergreen/container_registry_login.sh"
+
+  "build and push module images": &build_and_push_module_images
+    command: subprocess.exec
+    params:
+      binary: bash
+      add_expansions_to_env: true # needed to get the AWS secrets from ec2.assume_role
+      args:
+        - "./src/evergreen/build_and_push_module_images.sh"

etc/evergreen_yml_components/tasks/compile_tasks_shared.yml

@@ -212,6 +212,8 @@ tasks:
             - "src/src/**.yml"
             - "src/src/mongo/client/sdam/json_tests/sdam_tests/**"
             - "src/src/mongo/client/sdam/json_tests/server_selection_tests/**"
+            - "src/src/mongo/db/modules/atlas/evergreen/**"
+            - "src/src/mongo/db/modules/atlas/jstests/**"
             - "src/src/mongo/db/modules/enterprise/docs/**"
             - "src/src/mongo/db/modules/enterprise/jstests/**"
             - "src/src/mongo/db/modules/subscription/jstests/**"

etc/evergreen_yml_components/tasks/misc_tasks.yml

@@ -431,7 +431,13 @@ tasks:
 
   - <<: *run_jepsen_template
     name: jepsen_config_fuzzer_list-append
-    tags: ["assigned_to_jira_team_server_repl", "experimental", "jepsen_docker"]
+    tags:
+      [
+        "assigned_to_jira_team_server_repl",
+        "experimental",
+        "jepsen_docker",
+        "uses_docker",
+      ]
     commands:
       - func: "do setup"
       - func: "do jepsen docker setup"
@@ -514,7 +520,13 @@ tasks:
 
   - <<: *run_jepsen_template
     name: jepsen_list-append
-    tags: ["assigned_to_jira_team_server_repl", "experimental", "jepsen_docker"]
+    tags:
+      [
+        "assigned_to_jira_team_server_repl",
+        "experimental",
+        "jepsen_docker",
+        "uses_docker",
+      ]
     commands:
       - func: "do setup"
       - func: "do jepsen docker setup"
@@ -689,11 +701,7 @@ tasks:
 
   # Check that the mutational fuzzer can parse JS files modified in a patch build.
   - name: lint_fuzzer_sanity_patch
-    tags:
+    tags: ["assigned_to_jira_team_devprod_correctness", "experimental"]
-      [
-        "assigned_to_jira_team_devprod_correctness",
-        "development_critical_single_variant",
-      ]
     patch_only: true
     commands:
       - command: manifest.load
@@ -1517,6 +1525,22 @@ tasks:
     commands:
       - func: "generate smoke test tasks"
 
+  - name: push_mongod_to_ecr
+    tags: ["assigned_to_jira_team_disag_mongod"]
+    depends_on:
+      - name: package
+    commands:
+      - command: manifest.load
+      - func: "git get project and add git tag"
+      - func: "f_expansions_write"
+      - func: "set up venv"
+      - func: "fetch dist tarball"
+      - func: "extract binaries"
+      - command: ec2.assume_role
+        params:
+          role_arn: "${disagg_storage_ecr_arn}"
+      - func: "build and push module images"
+
   - name: selinux_rhel8_enterprise
     tags: ["assigned_to_jira_team_server_security", "experimental"]
     depends_on:

etc/evergreen_yml_components/tasks/resmoke/server_divisions/clusters_and_integrations/tasks.yml

@@ -1962,14 +1962,15 @@ tasks:
 
   - <<: *jstestfuzz_template
     name: resharding_timeseries_fuzzer_gen
-    tags:
+    tags: [
-      [
         "assigned_to_jira_team_server_cluster_scalability",
         "default",
         "feature_flag_guarded",
         "random_name",
         "require_npm",
         "requires_all_feature_flags",
+        # TODO SERVER-109849: Remove this tag.
+        "incompatible_disaggregated_storage",
       ]
     commands:
       - func: "generate resmoke tasks"

etc/evergreen_yml_components/tasks/resmoke/server_divisions/durable_transactions_and_availability/tasks.yml

@@ -667,12 +667,13 @@ tasks:
   - <<: *jstestfuzz_template
     name: initial_sync_fuzzer_sanity_patch_gen
     patch_only: true
-    tags:
+    tags: [
-      [
         "assigned_to_jira_team_server_repl",
         "default",
         "require_npm",
         "random_name",
+        # TODO SERVER-109849: Remove this tag.
+        "incompatible_disaggregated_storage",
       ]
     commands:
       - func: "generate resmoke tasks"
@@ -1551,13 +1552,14 @@ tasks:
   - <<: *jstestfuzz_template
     name: rollback_fuzzer_sanity_patch_gen
     patch_only: true
-    tags:
+    tags: [
-      [
         "assigned_to_jira_team_server_repl",
         "default",
         "rollbackfuzzer",
         "require_npm",
         "random_name",
+        # TODO SERVER-109849: Remove this tag.
+        "incompatible_disaggregated_storage",
       ]
     commands:
       - func: "generate resmoke tasks"
@@ -2341,3 +2343,30 @@ tasks:
       - func: "generate resmoke tasks"
         vars:
           suite: v1index_jscore_passthrough
+
+  ################################################
+  #           Disagg Storage tasks               #
+  ################################################
+
+  - <<: *gen_task_template
+    name: disagg_storage_gen
+    tags:
+      [
+        "assigned_to_jira_team_server_disagg",
+        "default",
+        "large",
+        "clustered_collections",
+        "uses_docker",
+      ]
+    commands:
+      - func: "generate resmoke tasks"
+        vars:
+          suite: disagg_storage
+          use_large_distro: "true"
+
+  - <<: *task_template
+    name: disagg_repl_jscore_passthrough
+    tags: ["assigned_to_jira_team_server_disagg", "default", "uses_docker"]
+    commands:
+      - func: "do setup"
+      - func: "run tests"

etc/evergreen_yml_components/variants/amazon/test_dev.yml

@@ -85,7 +85,7 @@ buildvariants:
 
   - <<: *linux-arm64-dynamic-compile-params
     name: &amazon-linux2023-arm64-static-compile amazon-linux2023-arm64-static-compile
-    display_name: "! Amazon Linux 2023 arm64 Enterprise Compile"
+    display_name: "! Amazon Linux 2023 arm64 Atlas Compile"
     tags: ["required", "bazel_check", "forbid_tasks_tagged_with_experimental"]
     expansions:
       <<: *linux-arm64-static-enterprise-compile-expansions
@@ -104,7 +104,6 @@ buildvariants:
       # since it's running on a c6g.16xlarge
       bazel_compile_flags: >-
         --define=MONGO_DISTMOD=amazon2023
-        --//bazel/config:build_otel=True
         --remote_execution_priority=3
         --jobs=1600
         --build_atlas=True
@@ -169,8 +168,8 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
         distros:
           - amazon2023-arm64-atlas-latest-large
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
         distros:
           - amazon2023-arm64-atlas-latest-large
       - name: .fuzzer_deterministic

etc/evergreen_yml_components/variants/amazon/test_dev_master_branch_only.yml

@@ -168,10 +168,16 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only
         distros:
           - amazon2023-arm64-atlas-latest-large
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only
         distros:
           - amazon2023-arm64-atlas-latest-large
+      - name: .default !.requires_large_host .uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only
+        distros:
+          - amazon2023-arm64-latest-small
+      - name: .default .requires_large_host .uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only
+        distros:
+          - amazon2023-arm64-latest-large
       - name: .fuzzer_deterministic
 
   - <<: *enterprise-amazon-linux2023-arm64-all-feature-flags-template
@@ -193,8 +199,8 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.requires_all_feature_flags
         distros:
           - amazon2023-arm64-latest-large
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.requires_all_feature_flags
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.requires_all_feature_flags
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.requires_all_feature_flags
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.requires_all_feature_flags
         distros:
           - amazon2023-arm64-latest-large
       - name: .fuzzer_deterministic
@@ -219,8 +225,8 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags !.multiversion !.suggested_excluding_required__for_devprod_mitigation_only
         distros:
           - amazon2023-arm64-atlas-latest-large
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags !.multiversion !.suggested_excluding_required__for_devprod_mitigation_only
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags !.multiversion !.suggested_excluding_required__for_devprod_mitigation_only
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags !.multiversion !.suggested_excluding_required__for_devprod_mitigation_only
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags !.multiversion !.suggested_excluding_required__for_devprod_mitigation_only
         distros:
           - amazon2023-arm64-atlas-latest-large
       - name: .fuzzer_deterministic
@@ -524,8 +530,8 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.multiversion
         distros:
           - amazon2023-arm64-latest-large
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.multiversion
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.multiversion
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.multiversion
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.suggested_excluding_required__for_devprod_mitigation_only !.multiversion
         distros:
           - amazon2023-arm64-latest-large
       - name: .fuzzer_deterministic !.multiversion
@@ -548,8 +554,8 @@ buildvariants:
   #     - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.multiversion !.serverless !.exclude_when_record_ids_replicated
   #       distros:
   #         - amazon2023-arm64-atlas-latest-large
-  #     - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.multiversion !.serverless !.exclude_when_record_ids_replicated
+  #     - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.multiversion !.serverless !.exclude_when_record_ids_replicated
-  #     - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.multiversion !.serverless !.exclude_when_record_ids_replicated
+  #     - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags !.multiversion !.serverless !.exclude_when_record_ids_replicated
   #       distros:
   #         - amazon2023-arm64-atlas-latest-large
   #   expansions:

etc/evergreen_yml_components/variants/rhel/test_dev.yml

@@ -129,8 +129,8 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
         distros:
           - rhel8.8-medium
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.requires_all_feature_flags
         distros:
           - rhel8.8-medium
 

etc/evergreen_yml_components/variants/rhel/test_dev_master_branch_only.yml

@@ -303,8 +303,8 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags
         distros:
           - rhel8.8-medium
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_all_feature_flags
         distros:
           - rhel8.8-medium
 

etc/evergreen_yml_components/variants/sanitizer/test_dev_master_branch_only.yml

@@ -226,11 +226,11 @@ buildvariants:
       - name: .release_critical .requires_large_host_debug_mode !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
         distros:
           - *enterprise-rhel-8-64-bit-dynamic-debug-mode-large-distro-name
-      - name: .default !.requires_large_host !.requires_large_host_debug_mode !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
+      - name: .default !.requires_large_host !.requires_large_host_debug_mode !.uses_docker !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
-      - name: .default .requires_large_host !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
+      - name: .default .requires_large_host !.uses_docker !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
         distros:
           - *enterprise-rhel-8-64-bit-dynamic-debug-mode-large-distro-name
-      - name: .default .requires_large_host_debug_mode !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
+      - name: .default .requires_large_host_debug_mode !.uses_docker !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
         distros:
           - *enterprise-rhel-8-64-bit-dynamic-debug-mode-large-distro-name
       - name: .non_deterministic !.requires_large_host !.requires_large_host_debug_mode !.incompatible_development_variant !.incompatible_debug_mode !.incompatible_system_allocator !.requires_all_feature_flags
@@ -480,8 +480,8 @@ buildvariants:
       - name: .release_critical .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_aubsan !.incompatible_system_allocator !.incompatible_all_feature_flags
         distros:
           - rhel8.8-xlarge
-      - name: .default !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_aubsan !.incompatible_system_allocator !.incompatible_all_feature_flags
+      - name: .default !.requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_aubsan !.incompatible_system_allocator !.incompatible_all_feature_flags
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_aubsan !.incompatible_system_allocator !.incompatible_all_feature_flags
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_aubsan !.incompatible_system_allocator !.incompatible_all_feature_flags
         distros:
           - rhel8.8-xlarge
       - name: .non_deterministic !.requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_aubsan !.incompatible_system_allocator !.incompatible_all_feature_flags
@@ -583,11 +583,11 @@ buildvariants:
       - name: .release_critical .requires_large_host_tsan !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags
         distros:
           - *enterprise-rhel8-debug-tsan-large-distro-name
-      - name: .default !.requires_large_host !.requires_large_host_tsan !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags
+      - name: .default !.requires_large_host !.requires_large_host_tsan !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags
-      - name: .default .requires_large_host !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags
+      - name: .default .requires_large_host !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags
         distros:
           - *enterprise-rhel8-debug-tsan-large-distro-name
-      - name: .default .requires_large_host_tsan !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags
+      - name: .default .requires_large_host_tsan !.uses_docker !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags
         distros:
           - *enterprise-rhel8-debug-tsan-large-distro-name
       - name: .non_deterministic !.requires_large_host !.requires_large_host_tsan !.requires_compile_variant !.incompatible_development_variant !.incompatible_tsan !.incompatible_system_allocator !.incompatible_all_feature_flags

evergreen/build_and_push_module_images.sh

@@ -0,0 +1,7 @@
+set -e
+
+for dir in ./src/src/mongo/db/modules/*; do
+    if test -f $dir/evergreen/build_and_push_images.sh; then
+        bash $dir/evergreen/build_and_push_images.sh
+    fi
+done

evergreen/fetch_module_images.sh

@@ -0,0 +1,7 @@
+set -e
+
+for dir in ./src/src/mongo/db/modules/*; do
+    if test -f $dir/evergreen/fetch_images.sh; then
+        bash $dir/evergreen/fetch_images.sh
+    fi
+done

jstests/libs/OWNERS.yml

@@ -87,3 +87,9 @@ filters:
   - "json_utils.js":
     approvers:
       - 10gen/query-integration-extensions
+  - "replicated_ident_utils.js":
+    approvers:
+      - 10gen/server-storage-engine-integration
+  - "replicated_record_ids_utils.js":
+    approvers:
+      - 10gen/server-storage-engine-integration

jstests/libs/collection_write_path/replicated_record_ids_utils.js

@@ -2,7 +2,7 @@
 // documents, otherwise.
 import {ReplSetTest} from "jstests/libs/replsettest.js";
 
-function getShowRecordIdsCursor(node, dbName, replicatedCollName) {
+export function getShowRecordIdsCursor(node, dbName, replicatedCollName) {
     return node
         .getDB(dbName)
         [replicatedCollName].aggregate([{"$project": {"recordId": {"$meta": "recordId"}, "document": "$$ROOT"}}]);

jstests/libs/replicated_ident_utils.js

@@ -0,0 +1,86 @@
+/*
+ * Helpers for basic testing of replicated idents.
+ */
+function getOplog(node) {
+    return node.getDB("local").oplog.rs;
+}
+
+export function getSortedCatalogEntries(node, sortField = "ident") {
+    const adminDB = node.getDB("admin");
+    const isSystemProfile = {"name": "system.profile"};
+    const isLocal = {"db": "local"};
+    const match = {$nor: [isSystemProfile, isLocal]};
+    return adminDB.aggregate([{$listCatalog: {}}, {$match: match}, {$sort: {sortField: 1}}]).toArray();
+}
+/**
+ * Given catalog entries for 2 nodes, where catalog entries for both nodes must be sorted by the
+ * same field, validates that each entry has a matching 'ident'.
+ */
+export function assertMatchingCatalogIdents(node0CatalogIdents, node1CatalogIdents) {
+    jsTest.log(
+        `Asserting catalog entries for node0 ${tojson(node0CatalogIdents)} with node1 ${tojson(node1CatalogIdents)}`,
+    );
+    assert.eq(
+        node0CatalogIdents.length,
+        node1CatalogIdents.length,
+        `Expected nodes to have same number of entries. Entries for node0 ${tojson(
+            node0CatalogIdents,
+        )}, entries for node1 ${node1CatalogIdents}`,
+    );
+
+    const numCatalogEntries = node0CatalogIdents.length;
+    const entriesThatDontMatch = [];
+    for (let i = 0; i < numCatalogEntries; i++) {
+        const entryNode0 = node0CatalogIdents[i];
+        const entryNode1 = node1CatalogIdents[i];
+
+        if (bsonWoCompare(entryNode0, entryNode1) !== 0) {
+            // For visibility, collect all mismatched entries before failing.
+            entriesThatDontMatch.push([entryNode0, entryNode1]);
+            jsTest.log(
+                `Expected both nodes to have same entries. Node0 has ${tojson(
+                    entryNode0,
+                )}, Node1 has ${tojson(entryNode1)}`,
+            );
+        }
+    }
+
+    assert.eq(
+        0,
+        entriesThatDontMatch.length,
+        `Catalog entries for were expected to match, but don't. Entries that don't match ${tojson(
+            entriesThatDontMatch,
+        )}`,
+    );
+}
+
+// Validates that all 'create' collection oplog entries contain collection idents.
+export function assertCreateOplogEntriesContainIdents(node) {
+    const createOps = getOplog(node)
+        .find({"op": "c", "o.create": {$exists: true}})
+        .toArray();
+    jsTest.log("Create oplog entries on node " + node.port + " " + tojson(createOps));
+    assert.lt(0, createOps.length);
+    for (let op of createOps) {
+        assert(
+            op.hasOwnProperty("o2"),
+            `Expected to have 'o2' field present in ${tojson(
+                op,
+            )}. Dumping all create oplog entries ${tojson(createOps)}`,
+        );
+
+        const o2 = op["o2"];
+        assert(
+            o2.hasOwnProperty("ident"),
+            `Expected to find 'ident' property in 'o2' field of ${tojson(
+                op,
+            )}. Dumping all create oplog entries ${tojson(createOps)}`,
+        );
+        assert(
+            o2.hasOwnProperty("idIndexIdent"),
+            `Expected to find 'iddIndexIdent' property in 'o2' field of ${tojson(
+                op,
+            )}. Dumping all create oplog entries ${tojson(createOps)}`,
+        );
+    }
+}

jstests/libs/replsettest.js

@@ -1768,6 +1768,28 @@ export class ReplSetTest {
         });
     }
 
+    /**
+     * Runs replSetInitiate on the first node of the replica set.
+     *
+     * TODO (SERVER-109841): Replsetinitiate is currently a no-op command for disagg. Determine the
+     * next steps for this function if additional functionality is to be incorporated.
+     */
+    initiateForDisagg(cfg, initCmd) {
+        const startTime = new Date(); // Measure the execution time of this function.
+
+        // Blocks until there is a primary. We use a faster retry interval here since we expect the
+        // primary to be ready very soon. We also turn the failpoint off once we have a primary.
+        this.getPrimary(this.kDefaultTimeoutMS, 25 /* retryIntervalMS */);
+
+        jsTest.log(
+            "ReplSetTest initiateForDisagg took " +
+                (new Date() - startTime) +
+                "ms for " +
+                this.nodes.length +
+                " nodes.",
+        );
+    }
+
     /**
      * Steps up 'node' as primary and by default it waits for the stepped up node to become a
      * writable primary and waits for all nodes to reach the same optime before sending the
@@ -3589,7 +3611,7 @@ function _constructStartNewInstances(rst, opts) {
         rst._unbridgedPorts = Array.from({length: numNodes}, rst._allocatePortForNode);
         rst._unbridgedNodes = [];
     } else {
-        rst.ports = Array.from({length: numNodes}, rst._allocatePortForNode);
+        rst.ports = opts.ports || Array.from({length: numNodes}, rst._allocatePortForNode);
     }
 
     for (let i = 0; i < numNodes; i++) {

modules_poc/modules.yaml

@@ -530,6 +530,7 @@ encrypted_storage_engine:
     slack: server-security
     jira: Server Security
   files:
+    - src/mongo/db/modules/atlas/src/disagg_storage/encryption
     - src/mongo/db/modules/enterprise/src/encryptdb
 
 security:
@@ -854,8 +855,7 @@ disagg_storage:
     slack: disaggregated-storage-mongod
     jira: RSSD
   files:
-    - src/mongo/db/modules/atlas
+    - src/mongo/db/modules/atlas/src/disagg_storage
-    - src/mongo/db/disagg_storage
 
 storage_engine_api:
   meta:
@@ -992,6 +992,13 @@ installer:
   files:
     - src/mongo/installer/
 
+replicated_storage_service:
+  meta:
+    slack: server-replication
+    jira: Server Replication
+  files:
+    - src/mongo/db/rss
+
 replication:
   meta:
     slack: server-replication

sbom.json

@@ -914,6 +914,49 @@
       },
       "scope": "required"
     },
+    {
+      "supplier": {
+        "name": "Organization: github"
+      },
+      "name": "googletest",
+      "version": "1.17.0",
+      "licenses": [
+        {
+          "license": {
+            "id": "BSD-3-Clause"
+          }
+        }
+      ],
+      "purl": "pkg:github/googletest/googletest@v1.17.0",
+      "properties": [
+        {
+          "name": "internal:team_responsible",
+          "value": "Disaggregated Storage"
+        },
+        {
+          "name": "emits_persisted_data",
+          "value": "false"
+        },
+        {
+          "name": "info_link",
+          "value": "https://github.com/google/googletest"
+        },
+        {
+          "name": "import_script_path",
+          "value": "src/third_party/googletest_restricted_for_disagg_only/scripts/import.sh"
+        }
+      ],
+      "type": "library",
+      "bom-ref": "e57f94bd-b0b1-4e47-912e-c690a01e4f95",
+      "evidence": {
+        "occurrences": [
+          {
+            "location": "src/third_party/googletest_restricted_for_disagg_only"
+          }
+        ]
+      },
+      "scope": "required"
+    },
     {
       "type": "library",
       "bom-ref": "pkg:github/gperftools/gperftools@gperftools-2.9.1",

src/mongo/BUILD.bazel

@@ -21,6 +21,11 @@ generate_config_header(
             "MONGO_CONFIG_OTEL": "1",
         },
         "//conditions:default": {},
+    }) | select({
+        "//bazel/config:build_atlas_enabled": {
+            "MONGO_CONFIG_DISAGG_STORAGE": "1",
+        },
+        "//conditions:default": {},
     }) | select({
         "//bazel/config:mutex_observation_enabled": {
             "MONGO_CONFIG_MUTEX_OBSERVATION": "1",

src/mongo/config.h.in

@@ -125,5 +125,8 @@
 // Defined if the build includes OpenTelemetry
 @mongo_config_otel@
 
+// Defined if the build includes disaggregated storage
+@mongo_config_disagg_storage@
+
 // Defined if the build includes mutex observation
 @mongo_config_mutex_observation@

src/mongo/db/BUILD.bazel

@@ -3320,6 +3320,8 @@ mongo_cc_library(
         "//src/mongo/db/repl:serveronly_repl",
         "//src/mongo/db/repl:storage_interface_impl",
         "//src/mongo/db/repl:topology_coordinator",
+        "//src/mongo/db/rss:persistence_provider_impl",
+        "//src/mongo/db/rss:service_lifecycle_impl",
         "rw_concern_d",
         "//src/mongo/db/session:kill_sessions_local",
         "//src/mongo/db/session:service_liaison_mongod",
@@ -3433,6 +3435,11 @@ mongo_cc_library(
             "//src/mongo/db/modules/enterprise/src/kmip:kmip_configuration",
         ],
         "//conditions:default": [],
+    }) + select({
+        "//bazel/config:build_atlas_enabled": [
+            "//src/mongo/db/modules/atlas/src/disagg_storage/encryption:sls_log_encryption_manager",
+        ],
+        "//conditions:default": [],
     }),
 )
 
@@ -3530,6 +3537,7 @@ mongo_cc_library(
         "//src/mongo/db/repl:storage_interface_impl",
         "//src/mongo/db/repl:topology_coordinator",
         "//src/mongo/db/repl:wait_for_majority_service",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/s:query_analysis_writer",
         "//src/mongo/db/s:sessions_collection_config_server",
         "//src/mongo/db/s:sharding_commands_d",
@@ -3563,11 +3571,6 @@ mongo_cc_library(
             "//src/mongo/util/tracing_profiler",
         ],
         "//conditions:default": [],
-    }) + select({
-        "//bazel/config:build_atlas_required_settings": [
-            "//src/mongo/db/modules/atlas:atlas_only",
-        ],
-        "//conditions:default": [],
     }),
 )
 
@@ -3697,6 +3700,7 @@ mongo_cc_library(
         "//src/mongo/db/local_catalog:catalog_impl",
         "//src/mongo/db/op_observer",
         "//src/mongo/db/repl:replmocks",
+        "//src/mongo/db/rss:persistence_provider_impl",
         "//src/mongo/db/s:sharding_runtime_d",
         "//src/mongo/db/storage:storage_control",
         "//src/mongo/db/storage:storage_options",
@@ -4138,6 +4142,7 @@ mongo_cc_library(
         "//src/mongo/db/local_catalog:database_holder",
         "//src/mongo/db/op_observer",
         "//src/mongo/db/repl:replmocks",
+        "//src/mongo/db/rss:persistence_provider_impl",
         "//src/mongo/db/s:sharding_runtime_d",
         "//src/mongo/db/stats:top",
         "//src/mongo/db/storage:storage_control",
@@ -4369,6 +4374,7 @@ mongo_cc_benchmark(
         "//src/mongo/db/repl:repl_coordinator_impl",
         "//src/mongo/db/repl:serveronly_repl",
         "//src/mongo/db/repl:storage_interface_impl",
+        "//src/mongo/db/rss:persistence_provider_impl",
         "//src/mongo/db/s:sharding_runtime_d",
         "//src/mongo/db/storage:storage_control",
         "//src/mongo/db/storage/wiredtiger:storage_wiredtiger",

src/mongo/db/admission/flow_control.cpp

@@ -127,24 +127,62 @@ Timestamp getMedianAppliedTimestamp(const std::vector<repl::MemberData>& sortedM
     const int sustainerIdx = sortedMemberData.size() / 2;
     return sortedMemberData[sustainerIdx].getLastAppliedOpTime().getTimestamp();
 }
+}  // namespace
+
+namespace flow_control_details {
+ReplicationTimestampProvider::ReplicationTimestampProvider(repl::ReplicationCoordinator* replCoord)
+    : _replCoord(replCoord) {}
+
+Timestamp ReplicationTimestampProvider::getCurrSustainerTimestamp() const {
+    return getMedianAppliedTimestamp(_currMemberData);
+}
+
+Timestamp ReplicationTimestampProvider::getPrevSustainerTimestamp() const {
+    return getMedianAppliedTimestamp(_prevMemberData);
+}
+
+repl::TimestampAndWallTime ReplicationTimestampProvider::getTargetTimestampAndWallTime() const {
+    auto time = _replCoord->getLastCommittedOpTimeAndWallTime();
+    return {.timestamp = time.opTime.getTimestamp(), .wallTime = time.wallTime};
+}
+
+repl::TimestampAndWallTime ReplicationTimestampProvider::getLastWriteTimestampAndWallTime() const {
+    auto time = _replCoord->getMyLastAppliedOpTimeAndWallTime();
+    return {.timestamp = time.opTime.getTimestamp(), .wallTime = time.wallTime};
+}
+
+void ReplicationTimestampProvider::update() {
+    _prevMemberData = _currMemberData;
+    _currMemberData = _replCoord->getMemberData();
+
+    // Sort MemberData with the 0th index being the node with the lowest applied optime.
+    std::sort(_currMemberData.begin(),
+              _currMemberData.end(),
+              [](const repl::MemberData& left, const repl::MemberData& right) -> bool {
+                  return left.getLastAppliedOpTime() < right.getLastAppliedOpTime();
+              });
+}
+
+bool ReplicationTimestampProvider::flowControlUsable() const {
+    return _replCoord->canAcceptNonLocalWrites();
+}
 
 /**
  * Sanity checks whether the successive queries of topology data are comparable for doing a flow
  * control calculation. In particular, the number of members must be the same and the median
  * applier's timestamp must not go backwards.
  */
-bool sustainerAdvanced(const std::vector<repl::MemberData>& prevMemberData,
+bool ReplicationTimestampProvider::sustainerAdvanced() const {
-                       const std::vector<repl::MemberData>& currMemberData) {
+    if (_currMemberData.size() == 0 || _currMemberData.size() != _prevMemberData.size()) {
-    if (currMemberData.size() == 0 || currMemberData.size() != prevMemberData.size()) {
         LOGV2_WARNING(22223,
                       "Flow control detected a change in topology",
-                      "prevSize"_attr = prevMemberData.size(),
+                      "prevSize"_attr = _prevMemberData.size(),
-                      "currSize"_attr = currMemberData.size());
+                      "currSize"_attr = _currMemberData.size());
         return false;
     }
 
-    auto currSustainerAppliedTs = getMedianAppliedTimestamp(currMemberData);
+    auto currSustainerAppliedTs = getMedianAppliedTimestamp(_currMemberData);
-    auto prevSustainerAppliedTs = getMedianAppliedTimestamp(prevMemberData);
+    auto prevSustainerAppliedTs = getMedianAppliedTimestamp(_prevMemberData);
 
     if (currSustainerAppliedTs < prevSustainerAppliedTs) {
         LOGV2_WARNING(22224,
@@ -156,13 +194,42 @@ bool sustainerAdvanced(const std::vector<repl::MemberData>& prevMemberData,
 
     return true;
 }
-}  // namespace
+
+void ReplicationTimestampProvider::setCurrMemberData_forTest(
+    const std::vector<repl::MemberData>& memberData) {
+    _currMemberData = memberData;
+    std::sort(_currMemberData.begin(),
+              _currMemberData.end(),
+              [](const repl::MemberData& left, const repl::MemberData& right) -> bool {
+                  return left.getLastAppliedOpTime() < right.getLastAppliedOpTime();
+              });
+}
+
+void ReplicationTimestampProvider::setPrevMemberData_forTest(
+    const std::vector<repl::MemberData>& memberData) {
+    _prevMemberData = memberData;
+    std::sort(_prevMemberData.begin(),
+              _prevMemberData.end(),
+              [](const repl::MemberData& left, const repl::MemberData& right) -> bool {
+                  return left.getLastAppliedOpTime() < right.getLastAppliedOpTime();
+              });
+}
+
+}  // namespace flow_control_details
 
 FlowControl::FlowControl(repl::ReplicationCoordinator* replCoord)
-    : _replCoord(replCoord), _lastTimeSustainerAdvanced(Date_t::now()) {}
+    : _timestampProvider(
+          std::make_unique<flow_control_details::ReplicationTimestampProvider>(replCoord)),
+      _lastTimeSustainerAdvanced(Date_t::now()) {}
 
 FlowControl::FlowControl(ServiceContext* service, repl::ReplicationCoordinator* replCoord)
-    : _replCoord(replCoord), _lastTimeSustainerAdvanced(Date_t::now()) {
+    : FlowControl(service,
+                  std::make_unique<flow_control_details::ReplicationTimestampProvider>(replCoord)) {
+}
+
+FlowControl::FlowControl(ServiceContext* service,
+                         std::unique_ptr<TimestampProvider> timestampProvider)
+    : _timestampProvider(std::move(timestampProvider)), _lastTimeSustainerAdvanced(Date_t::now()) {
     // Initialize _lastTargetTicketsPermitted to maximum tickets to make sure flow control doesn't
     // cause a slow start on start up.
     FlowControlTicketholder::set(service, std::make_unique<FlowControlTicketholder>(kMaxTickets));
@@ -254,44 +321,26 @@ void FlowControl::disableUntil(Date_t deadline) {
     _disableUntil.store(deadline);
 }
 
-/**
+int FlowControl::_calculateNewTicketsForLag(const Timestamp& prevSustainerTimestamp,
- * Advance the `_*MemberData` fields and sort the new data by the element's last applied optime.
+                                            const Timestamp& currSustainerTimestamp,
- */
-void FlowControl::_updateTopologyData() {
-    _prevMemberData = _currMemberData;
-    _currMemberData = _replCoord->getMemberData();
-
-    // Sort MemberData with the 0th index being the node with the lowest applied optime.
-    std::sort(_currMemberData.begin(),
-              _currMemberData.end(),
-              [](const repl::MemberData& left, const repl::MemberData& right) -> bool {
-                  return left.getLastAppliedOpTime() < right.getLastAppliedOpTime();
-              });
-}
-
-int FlowControl::_calculateNewTicketsForLag(const std::vector<repl::MemberData>& prevMemberData,
-                                            const std::vector<repl::MemberData>& currMemberData,
                                             std::int64_t locksUsedLastPeriod,
                                             double locksPerOp,
                                             std::uint64_t lagMillis,
                                             std::uint64_t thresholdLagMillis) {
+    invariant(prevSustainerTimestamp <= currSustainerTimestamp,
+              fmt::format("PrevSustainer: {} CurrSustainer: {}",
+                          prevSustainerTimestamp.toString(),
+                          currSustainerTimestamp.toString()));
     invariant(lagMillis >= thresholdLagMillis);
 
-    const auto currSustainerAppliedTs = getMedianAppliedTimestamp(currMemberData);
-    const auto prevSustainerAppliedTs = getMedianAppliedTimestamp(prevMemberData);
-    invariant(prevSustainerAppliedTs <= currSustainerAppliedTs,
-              fmt::format("PrevSustainer: {} CurrSustainer: {}",
-                          prevSustainerAppliedTs.toString(),
-                          currSustainerAppliedTs.toString()));
-
     const std::int64_t sustainerAppliedCount =
-        _approximateOpsBetween(prevSustainerAppliedTs, currSustainerAppliedTs);
+        _approximateOpsBetween(prevSustainerTimestamp, currSustainerTimestamp);
     LOGV2_DEBUG(22218,
                 DEBUG_LOG_LEVEL,
-                " PrevApplied: {prevSustainerAppliedTs} CurrApplied: {currSustainerAppliedTs} "
+                " PrevApplied: {prevSustainerTimestamp} CurrApplied: {currSustainerTimestamp} "
                 "NumSustainerApplied: {sustainerAppliedCount}",
-                "prevSustainerAppliedTs"_attr = prevSustainerAppliedTs,
+                "prevSustainerTimestamp"_attr = prevSustainerTimestamp,
-                "currSustainerAppliedTs"_attr = currSustainerAppliedTs,
+                "currSustainerTimestamp"_attr = currSustainerTimestamp,
                 "sustainerAppliedCount"_attr = sustainerAppliedCount);
     if (sustainerAppliedCount > 0) {
         _lastTimeSustainerAdvanced = Date_t::now();
@@ -359,35 +408,35 @@ int FlowControl::getNumTickets(Date_t now) {
     }
 
     // Flow Control is only enabled on nodes that can accept writes.
-    const bool canAcceptWrites = _replCoord->canAcceptNonLocalWrites();
+    const bool flowControlUsable = _timestampProvider->flowControlUsable();
 
     if (auto sfp = flowControlTicketOverride.scoped(); MONGO_unlikely(sfp.isActive())) {
         int numTickets = sfp.getData().getIntField("numTickets");
-        if (numTickets > 0 && canAcceptWrites) {
+        if (numTickets > 0 && flowControlUsable) {
             return numTickets;
         }
     }
 
     // It's important to update the topology on each iteration.
-    _updateTopologyData();
+    _timestampProvider->update();
-    const repl::OpTimeAndWallTime myLastApplied = _replCoord->getMyLastAppliedOpTimeAndWallTime();
+    const auto lastWriteTime = _timestampProvider->getLastWriteTimestampAndWallTime();
-    const repl::OpTimeAndWallTime lastCommitted = _replCoord->getLastCommittedOpTimeAndWallTime();
+    const auto lastTargetTime = _timestampProvider->getTargetTimestampAndWallTime();
     const double locksPerOp = _getLocksPerOp();
     const std::int64_t locksUsedLastPeriod = _getLocksUsedLastPeriod();
 
-    if (gFlowControlEnabled.load() == false || canAcceptWrites == false || locksPerOp < 0.0) {
+    if (gFlowControlEnabled.load() == false || flowControlUsable == false || locksPerOp < 0.0) {
-        _trimSamples(std::min(lastCommitted.opTime.getTimestamp(),
+        _trimSamples(
-                              getMedianAppliedTimestamp(_prevMemberData)));
+            std::min(lastTargetTime.timestamp, _timestampProvider->getPrevSustainerTimestamp()));
         return kMaxTickets;
     }
 
     int ret = 0;
     const auto thresholdLagMillis = getThresholdLagMillis();
 
-    // Successive lastCommitted and lastApplied wall clock time recordings are not guaranteed to be
+    // Successive lastTargetTime and lastApplied wall clock time recordings are not guaranteed to be
     // monotonically increasing. Recordings that satisfy the following check result in a negative
     // value for lag, so ignore them.
-    const bool ignoreWallTimes = lastCommitted.wallTime > myLastApplied.wallTime;
+    const bool ignoreWallTimes = lastTargetTime.wallTime > lastWriteTime.wallTime;
 
     // _approximateOpsBetween will return -1 if the input timestamps are in the same "bucket".
     // This is an indication that there are very few ops between the two timestamps.
@@ -395,9 +444,8 @@ int FlowControl::getNumTickets(Date_t now) {
     // Don't let the no-op writer on idle systems fool the sophisticated "is the replica set
     // lagged" classifier.
     const bool isHealthy = !ignoreWallTimes &&
-        (getLagMillis(myLastApplied.wallTime, lastCommitted.wallTime) < thresholdLagMillis ||
+        (getLagMillis(lastWriteTime.wallTime, lastTargetTime.wallTime) < thresholdLagMillis ||
-         _approximateOpsBetween(lastCommitted.opTime.getTimestamp(),
+         _approximateOpsBetween(lastTargetTime.timestamp, lastWriteTime.timestamp) == -1);
-                                myLastApplied.opTime.getTimestamp()) == -1);
 
     if (isHealthy) {
         // The add/multiply technique is used to ensure ticket allocation can ramp up quickly,
@@ -412,16 +460,16 @@ int FlowControl::getNumTickets(Date_t now) {
             auto waitTime = curTimeMicros64() - _startWaitTime;
             _isLaggedTimeMicros.fetchAndAddRelaxed(waitTime);
         }
-    } else if (!ignoreWallTimes && sustainerAdvanced(_prevMemberData, _currMemberData)) {
+    } else if (!ignoreWallTimes && _timestampProvider->sustainerAdvanced()) {
         // Expected case where flow control has meaningful data from the last period to make a new
         // calculation.
-        ret =
+        ret = _calculateNewTicketsForLag(
-            _calculateNewTicketsForLag(_prevMemberData,
+            _timestampProvider->getPrevSustainerTimestamp(),
-                                       _currMemberData,
+            _timestampProvider->getCurrSustainerTimestamp(),
-                                       locksUsedLastPeriod,
+            locksUsedLastPeriod,
-                                       locksPerOp,
+            locksPerOp,
-                                       getLagMillis(myLastApplied.wallTime, lastCommitted.wallTime),
+            getLagMillis(lastWriteTime.wallTime, lastTargetTime.wallTime),
-                                       thresholdLagMillis);
+            thresholdLagMillis);
         if (!_isLagged.load()) {
             _isLagged.store(true);
             _isLaggedCount.fetchAndAddRelaxed(1);
@@ -443,9 +491,10 @@ int FlowControl::getNumTickets(Date_t now) {
                 DEBUG_LOG_LEVEL,
                 "FlowControl debug.",
                 "isLagged"_attr = (_isLagged.load() ? "true" : "false"),
-                "currlagMillis"_attr = getLagMillis(myLastApplied.wallTime, lastCommitted.wallTime),
+                "currlagMillis"_attr =
-                "opsLagged"_attr = _approximateOpsBetween(lastCommitted.opTime.getTimestamp(),
+                    getLagMillis(lastWriteTime.wallTime, lastTargetTime.wallTime),
-                                                          myLastApplied.opTime.getTimestamp()),
+                "opsLagged"_attr =
+                    _approximateOpsBetween(lastTargetTime.timestamp, lastWriteTime.timestamp),
                 "granting"_attr = ret,
                 "lastGranted"_attr = _lastTargetTicketsPermitted.load(),
                 "lastSustainerApplied"_attr = _lastSustainerAppliedCount.load(),
@@ -457,7 +506,7 @@ int FlowControl::getNumTickets(Date_t now) {
     _lastTargetTicketsPermitted.store(ret);
 
     _trimSamples(
-        std::min(lastCommitted.opTime.getTimestamp(), getMedianAppliedTimestamp(_prevMemberData)));
+        std::min(lastTargetTime.timestamp, _timestampProvider->getPrevSustainerTimestamp()));
 
     return ret;
 }

src/mongo/db/admission/flow_control.h

@@ -34,6 +34,7 @@
 #include "mongo/bson/timestamp.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/db/repl/member_data.h"
+#include "mongo/db/repl/optime.h"
 #include "mongo/db/repl/replication_coordinator.h"
 #include "mongo/db/repl/replication_coordinator_fwd.h"
 #include "mongo/db/service_context.h"
@@ -62,12 +63,66 @@ namespace mongo {
  */
 class FlowControl {
 public:
+    class TimestampProvider {
+    public:
+        virtual ~TimestampProvider() = default;
+        /**
+         * The sustainer timestamp is the timestamp which, if moved forward, will cause an
+         * advance in the target timestamp.  For replication, it is the median applied timestamp
+         * on all the relevant nodes.  We need to know this timestamp both for the current iteration
+         * and the previous iteration.
+         */
+        virtual Timestamp getCurrSustainerTimestamp() const = 0;
+        virtual Timestamp getPrevSustainerTimestamp() const = 0;
+
+        /**
+         * The target time is the time we are trying to throttle to.  For replication, it is the
+         * last committed time (majority snapshot time).
+         */
+        virtual repl::TimestampAndWallTime getTargetTimestampAndWallTime() const = 0;
+
+        /**
+         * The last write time is what we are trying to control.  For replication, it is
+         * the last applied time.
+         */
+        virtual repl::TimestampAndWallTime getLastWriteTimestampAndWallTime() const = 0;
+
+        /**
+         * Is flow control possible with this timestamp provider?  For replication,
+         * true if this is a primary and majority read concern is enabled.
+         */
+        virtual bool flowControlUsable() const = 0;
+
+        /**
+         * Are the previous and current updates compatible?  For replication,
+         * makes sure number of nodes is the same and the median node timestamp (the sustainer)
+         * has not gone backwards.
+         */
+        virtual bool sustainerAdvanced() const = 0;
+
+        /**
+         * Advance the `_*MemberData` fields and sort the new data by the element's last applied
+         * optime.
+         */
+        virtual void update() = 0;
+    };
+
     static constexpr int kMaxTickets = 1000 * 1000 * 1000;
 
+    /**
+     * Construct a flow control object based on a custom timestamp provider.
+     * Takes ownership of the timestamp provider.
+     */
+    FlowControl(ServiceContext* service, std::unique_ptr<TimestampProvider> timestampProvider);
+
+    /**
+     * Construct a replication-based flow control object.
+     */
     FlowControl(ServiceContext* service, repl::ReplicationCoordinator* replCoord);
 
     /**
-     * Construct a flow control object without adding a periodic job runner for testing.
+     * Construct a replication-based flow control object without adding a periodic job runner for
+     * testing.
      */
     FlowControl(repl::ReplicationCoordinator* replCoord);
 
@@ -122,13 +177,13 @@ public:
 
     std::int64_t _approximateOpsBetween(Timestamp prevTs, Timestamp currTs);
 
-    void _updateTopologyData();
+    int _calculateNewTicketsForLag(const Timestamp& prevSustainerTimestamp,
-    int _calculateNewTicketsForLag(const std::vector<repl::MemberData>& prevMemberData,
+                                   const Timestamp& currSustainerTimestamp,
-                                   const std::vector<repl::MemberData>& currMemberData,
                                    std::int64_t locksUsedLastPeriod,
                                    double locksPerOp,
                                    std::uint64_t lagMillis,
                                    std::uint64_t thresholdLagMillis);
+
     void _trimSamples(Timestamp trimSamplesTo);
 
     // Sample of (timestamp, ops, lock acquisitions) where ops and lock acquisitions are
@@ -139,7 +194,7 @@ public:
     }
 
 private:
-    repl::ReplicationCoordinator* _replCoord;
+    std::unique_ptr<TimestampProvider> _timestampProvider;
 
     // These values are updated with each flow control computation and are also surfaced in server
     // status.
@@ -161,9 +216,6 @@ private:
 
     std::int64_t _lastPollLockAcquisitions = 0;
 
-    std::vector<repl::MemberData> _currMemberData;
-    std::vector<repl::MemberData> _prevMemberData;
-
     Date_t _lastTimeSustainerAdvanced;
 
     // This value is used for calculating server status metrics.
@@ -172,4 +224,27 @@ private:
     PeriodicJobAnchor _jobAnchor;
 };
 
+namespace flow_control_details {
+class ReplicationTimestampProvider final : public FlowControl::TimestampProvider {
+public:
+    explicit ReplicationTimestampProvider(repl::ReplicationCoordinator* replCoord);
+    Timestamp getCurrSustainerTimestamp() const final;
+    Timestamp getPrevSustainerTimestamp() const final;
+    repl::TimestampAndWallTime getTargetTimestampAndWallTime() const final;
+    repl::TimestampAndWallTime getLastWriteTimestampAndWallTime() const final;
+    bool flowControlUsable() const final;
+    bool sustainerAdvanced() const final;
+    void update() final;
+
+    void setCurrMemberData_forTest(const std::vector<repl::MemberData>& memberData);
+    void setPrevMemberData_forTest(const std::vector<repl::MemberData>& memberData);
+
+private:
+    repl::ReplicationCoordinator* _replCoord;
+    std::vector<repl::MemberData> _currMemberData;
+    std::vector<repl::MemberData> _prevMemberData;
+};
+
+}  // namespace flow_control_details
+
 }  // namespace mongo

src/mongo/db/admission/flow_control_test.cpp

@@ -241,6 +241,15 @@ TEST_F(FlowControlTest, CalculatingTickets) {
     currMemberData.emplace_back(constructMemberData(Timestamp(2000)));
     currMemberData.emplace_back(constructMemberData(Timestamp(3000)));
 
+    flow_control_details::ReplicationTimestampProvider timestampProvider(replCoordMock);
+    timestampProvider.setPrevMemberData_forTest(prevMemberData);
+    timestampProvider.setCurrMemberData_forTest(currMemberData);
+    auto prevSustainerTimestamp = timestampProvider.getPrevSustainerTimestamp();
+    auto currSustainerTimestamp = timestampProvider.getCurrSustainerTimestamp();
+
+    ASSERT_EQ(Timestamp(1000), prevSustainerTimestamp);
+    ASSERT_EQ(Timestamp(2000), currSustainerTimestamp);
+
     // Construct samples where Timestamp X maps to operation number X.
     for (int ts = 1; ts <= 3000; ++ts) {
         flowControl->sample(Timestamp(ts), 1);
@@ -251,8 +260,8 @@ TEST_F(FlowControlTest, CalculatingTickets) {
     const std::uint64_t thresholdLag = 1;
     const std::uint64_t currLag = thresholdLag;
     ASSERT_EQ(1900,
-              flowControl->_calculateNewTicketsForLag(prevMemberData,
+              flowControl->_calculateNewTicketsForLag(prevSustainerTimestamp,
-                                                      currMemberData,
+                                                      currSustainerTimestamp,
                                                       locksUsedLastPeriod,
                                                       locksPerOp,
                                                       currLag,

src/mongo/db/auth/auth_op_observer.cpp

@@ -39,6 +39,7 @@
 #include "mongo/db/op_observer/op_observer_util.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/db/repl/oplog_entry.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/util/assert_util.h"
 #include "mongo/util/decorable.h"
 #include "mongo/util/namespace_string_util.h"
@@ -123,7 +124,9 @@ void AuthOpObserver::onCreateCollection(
 
     BSONObj o2;
     if (createCollCatalogIdentifier.has_value() &&
-        shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx))) {
+        shouldReplicateLocalCatalogIdentifers(
+            rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+            VersionContext::getDecoration(opCtx))) {
         o2 = repl::MutableOplogEntry::makeCreateCollObject2(
             createCollCatalogIdentifier->catalogId,
             createCollCatalogIdentifier->ident,

src/mongo/db/commands/BUILD.bazel

@@ -1471,6 +1471,7 @@ mongo_cc_library(
         "//src/mongo/db/repl:repl_server_parameters",
         "//src/mongo/db/repl:replica_set_messages",
         "//src/mongo/db/repl/dbcheck",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/s:sharding_catalog_manager",
         "//src/mongo/db/s:sharding_commands_d",
         "//src/mongo/db/s:transaction_coordinator",

src/mongo/db/commands/shutdown_d.cpp

@@ -35,6 +35,7 @@
 #include "mongo/db/index_builds/index_builds_coordinator.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/s/transaction_coordinator_service.h"
 #include "mongo/logv2/log.h"
 #include "mongo/platform/compiler.h"
@@ -71,7 +72,10 @@ Status stepDownForShutdown(OperationContext* opCtx,
 
             // Specify a high freeze time, so that if there is a stall during shut down, the node
             // does not run for election.
-            replCoord->stepDown(opCtx, false /* force */, waitTime, Days(1));
+            auto& rss = rss::ReplicatedStorageService::get(opCtx);
+            if (rss.getPersistenceProvider().shouldStepDownForShutdown()) {
+                replCoord->stepDown(opCtx, false /* force */, waitTime, Days(1));
+            }
 
             if (MONGO_unlikely(hangInShutdownAfterStepdown.shouldFail())) {
                 LOGV2(4695100, "hangInShutdownAfterStepdown failpoint enabled");

src/mongo/db/disagg_storage/BUILD.bazel

@@ -1,31 +0,0 @@
-load("//bazel:mongo_src_rules.bzl", "idl_generator", "mongo_cc_benchmark", "mongo_cc_library", "mongo_cc_unit_test")
-
-package(default_visibility = ["//visibility:public"])
-
-exports_files(
-    glob([
-        "*.h",
-        "*.cpp",
-    ]),
-)
-
-idl_generator(
-    name = "server_parameters_gen",
-    src = "server_parameters.idl",
-    deps = [
-        "//src/mongo/db:basic_types_gen",
-    ],
-)
-
-mongo_cc_library(
-    name = "server_parameters",
-    srcs = [
-        "server_parameters_gen",
-    ],
-    hdrs = [
-    ],
-    deps = [
-        "//src/mongo/db:server_base",
-        "//src/mongo/idl:idl_parser",
-    ],
-)

src/mongo/db/disagg_storage/OWNERS.yml

@@ -1,5 +0,0 @@
-version: 1.0.0
-filters:
-  - "*":
-    approvers:
-      - 10gen/server-disagg-storage

src/mongo/db/disagg_storage/server_parameters.idl

@@ -1,45 +0,0 @@
-# Copyright (C) 2025-present MongoDB, Inc.
-#
-# This program is free software: you can redistribute it and/or modify
-# it under the terms of the Server Side Public License, version 1,
-# as published by MongoDB, Inc.
-#
-# This program is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-# Server Side Public License for more details.
-#
-# You should have received a copy of the Server Side Public License
-# along with this program. If not, see
-# <http://www.mongodb.com/licensing/server-side-public-license>.
-#
-# As a special exception, the copyright holders give permission to link the
-# code of portions of this program with the OpenSSL library under certain
-# conditions as described in each individual source file and distribute
-# linked combinations including the program with the OpenSSL library. You
-# must comply with the Server Side Public License in all respects for
-# all of the code used other than as permitted herein. If you modify file(s)
-# with this exception, you may extend this exception to your version of the
-# file(s), but you are not obligated to do so. If you do not wish to do so,
-# delete this exception statement from your version. If you delete this
-# exception statement from all source files in the program, then also delete
-# it in the license file.
-#
-
-# server setParameters for disaggregated storage
-
-global:
-    cpp_namespace: "mongo::disagg"
-
-imports:
-    - "mongo/db/basic_types.idl"
-
-server_parameters:
-    disaggregatedStorageEnabled:
-        description: >-
-            Set this to run the server as a compute node in a disaggregated storage cluster.
-        set_at: startup
-        cpp_vartype: bool
-        cpp_varname: gDisaggregatedStorageEnabled
-        default: false
-        redact: false

src/mongo/db/exec/scoped_timer.h

@@ -158,7 +158,7 @@ private:
     X(abortAllTransactions)                     \
     X(joinLogicalSessionCache)                  \
     X(shutDownCursorManager)                    \
-    X(shutDownSLSStateMachine)                  \
+    X(shutDownStateRequiredForStorageAccess)    \
     /* For magic restore: */                    \
     X(magicRestoreToolTotal)                    \
     X(readMagicRestoreConfig)                   \

src/mongo/db/ftdc/ftdc_system_stats_linux.cpp

@@ -33,20 +33,31 @@
 #include "mongo/db/ftdc/collector.h"
 #include "mongo/db/ftdc/controller.h"
 #include "mongo/db/ftdc/ftdc_system_stats.h"
+#include "mongo/logv2/log.h"
 #include "mongo/util/errno_util.h"
 #include "mongo/util/functional.h"
 #include "mongo/util/processinfo.h"
 #include "mongo/util/procparser.h"
 
 #include <cstdint>
+#include <iostream>
 #include <memory>
+#include <set>
 #include <string>
 #include <utility>
 #include <vector>
 
+#include <ifaddrs.h>
+
+#include <linux/ethtool.h>
+#include <linux/if.h>
+#include <linux/sockios.h>
+#include <sys/ioctl.h>
 #include <sys/prctl.h>
 #include <sys/resource.h>
 
+#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kFTDC
+
 namespace mongo {
 
 namespace {
@@ -102,6 +113,188 @@ static const std::map<StringData, std::set<StringData>> kSockstatKeys{
     {"TCP"_sd, {"inuse"_sd, "orphan"_sd, "tw"_sd, "alloc"_sd}},
 };
 
+/**
+ * Class to gather NIC stats by emulating ethtool -S functionality by using the ioctl SIOCETHTOOL.
+ */
+class EthTool {
+public:
+    static std::unique_ptr<EthTool> create(StringData interface) {
+        int fd = socket(AF_INET, SOCK_DGRAM, 0);
+        if (fd == -1) {
+            auto ec = lastPosixError();
+            LOGV2_WARNING(
+                10985539, "Ethtool socket allocation failed", "error"_attr = errorMessage(ec));
+            return nullptr;
+        }
+
+        auto ethtool = std::unique_ptr<EthTool>(new EthTool(interface, fd));
+        auto drvinfo = ethtool->get_info();
+
+        // Some Linux interfaces cannot be found by ethtool IOCTL.
+        // Some Linux interfaces have no stats (i.e. the "bridge" driver used by containers).
+        if (!drvinfo.has_value() || drvinfo->n_stats == 0) {
+            LOGV2_WARNING(10985540,
+                          "Skipping Ethtool stats collection for interface",
+                          "interface"_attr = interface);
+            return nullptr;
+        }
+
+        return ethtool;
+    }
+
+    ~EthTool() {
+        free(_gstrings);
+
+        close(_fd);
+    }
+
+    // Get a list of all non-loopback interfaces for the machine
+    static std::vector<std::string> interface_names() {
+        struct ifaddrs* ifaddr;
+
+        if (getifaddrs(&ifaddr) == -1) {
+            auto ec = lastPosixError();
+            uasserted(10985538, fmt::format("getifaddrs failed: {}", errorMessage(ec)));
+        }
+        ON_BLOCK_EXIT([&] { freeifaddrs(ifaddr); });
+
+        std::set<std::string> names;
+        for (ifaddrs* ifa = ifaddr; ifa != NULL; ifa = ifa->ifa_next) {
+            if (ifa->ifa_addr == NULL) {
+                continue;
+            }
+
+            if ((ifa->ifa_flags & IFF_LOOPBACK) == IFF_LOOPBACK) {
+                continue;
+            }
+
+            names.insert(ifa->ifa_name);
+        }
+
+        std::vector<std::string> vec;
+        std::copy(names.begin(), names.end(), std::back_inserter(vec));
+        return vec;
+    }
+
+    // Get a list of stats names for a given interface
+    std::vector<StringData>& get_strings() {
+        if (!_names.has_value()) {
+            auto drvinfo = get_info();
+            _get_strings(drvinfo->n_stats);
+        }
+
+        return _names.get();
+    }
+
+    // Get a list of stats for a given interface
+    std::vector<uint64_t> get_stats() {
+        if (!_names.has_value()) {
+            return std::vector<uint64_t>();
+        }
+
+        return _get_stats(_names->size());
+    }
+
+    // Get a some basic information about the interface
+    boost::optional<ethtool_drvinfo> get_info() {
+        ethtool_drvinfo drvinfo;
+        memset(&drvinfo, 0, sizeof(drvinfo));
+        drvinfo.cmd = ETHTOOL_GDRVINFO;
+
+        if (_ioctlNoThrow("drvinfo", &drvinfo)) {
+            return boost::none;
+        }
+
+        return boost::optional<ethtool_drvinfo>(drvinfo);
+    }
+
+    // Name of the interface this class monitors
+    StringData name() const {
+        return _interface;
+    }
+
+private:
+    explicit EthTool(StringData interface, int fd) : _fd(fd), _interface(std::string(interface)) {}
+
+    void _get_strings(size_t count) {
+        _gstrings = static_cast<ethtool_gstrings*>(
+            calloc(1, sizeof(ethtool_gstrings) + count * ETH_GSTRING_LEN));
+
+        _gstrings->cmd = ETHTOOL_GSTRINGS;
+        _gstrings->string_set = ETH_SS_STATS;
+        _gstrings->len = count;
+
+        _names.emplace(std::vector<StringData>());
+
+        if (_ioctlNoThrow("get_strings", _gstrings)) {
+            return;
+        }
+
+        char* ptr = reinterpret_cast<char*>(_gstrings) + sizeof(ethtool_gstrings);
+        for (size_t i = 0; i < count; i++) {
+            auto s = StringData(ptr);
+
+            _names->push_back(s);
+
+            ptr += ETH_GSTRING_LEN;
+        }
+    }
+
+    std::vector<uint64_t> _get_stats(size_t count) {
+        std::vector<char> stats_buf(sizeof(ethtool_stats) + count * 8,
+                                    0); /* 8 is the number specfied in ethtool.h */
+
+        ethtool_stats* stats = reinterpret_cast<ethtool_stats*>(stats_buf.data());
+        stats->cmd = ETHTOOL_GSTATS;
+        stats->n_stats = count;
+
+        if (_ioctlNoThrow("get_stats", stats)) {
+            return std::vector<uint64_t>();
+        }
+
+        char* ptr = reinterpret_cast<char*>(stats) + sizeof(ethtool_stats);
+
+        std::vector<uint64_t> stats_vec(ptr, ptr + count * 8);
+
+        return stats_vec;
+    }
+
+    // Returns non-zero on error
+    int _ioctlNoThrow(StringData name, void* cmd) {
+        ifreq ifr;
+
+        strcpy(ifr.ifr_name, _interface.c_str());
+        ifr.ifr_data = cmd;
+
+        auto ret = ioctl(_fd, SIOCETHTOOL, &ifr);
+
+        if (MONGO_unlikely(ret) && !_warningLogged) {
+            auto ec = lastPosixError();
+            _warningLogged = true;
+
+            LOGV2_WARNING(10985553,
+                          "Failed to get strings for ethtool",
+                          "interface"_attr = _interface,
+                          "name"_attr = name,
+                          "error"_attr = errorMessage(ec));
+        }
+
+        return ret;
+    }
+
+private:
+    int _fd;
+
+    ethtool_gstrings* _gstrings{nullptr};
+
+    boost::optional<std::vector<StringData>> _names;
+
+    std::string _interface;
+
+    bool _warningLogged{false};
+};
+
+
 /**
  *  Collect metrics from the Linux /proc file system.
  */
@@ -111,6 +304,16 @@ public:
         for (const auto& disk : _disks) {
             _disksStringData.emplace_back(disk);
         }
+
+        auto interfaces = EthTool::interface_names();
+
+        _ethtools.reserve(interfaces.size());
+        for (const auto& ifn : interfaces) {
+            auto nic = EthTool::create(ifn);
+            if (nic) {
+                _ethtools.push_back(std::move(nic));
+            }
+        }
     }
 
     void collect(OperationContext* opCtx, BSONObjBuilder& builder) override {
@@ -219,6 +422,29 @@ public:
                 &subObjBuilder);
             subObjBuilder.doneFast();
         }
+
+        {
+            BSONObjBuilder subObjBuilder(builder.subobjStart("ethtool"_sd));
+
+            for (auto& tool : _ethtools) {
+                BSONObjBuilder subNICBuilder(subObjBuilder.subobjStart(tool->name()));
+
+                auto names = tool->get_strings();
+                if (names.empty()) {
+                    continue;
+                }
+
+                auto stats = tool->get_stats();
+                if (stats.empty()) {
+                    continue;
+                }
+                invariant(stats.size() >= names.size());
+
+                for (size_t i = 0; i < names.size(); i++) {
+                    subNICBuilder.append(names[i], static_cast<long long>(stats[i]));
+                }
+            }
+        }
     }
 
 private:
@@ -227,6 +453,8 @@ private:
 
     // List of physical disks to collect stats from as StringData to pass to parseProcDiskStatsFile.
     std::vector<StringData> _disksStringData;
+
+    std::vector<std::unique_ptr<EthTool>> _ethtools;
 };
 
 class SimpleFunctionCollector final : public FTDCCollectorInterface {

src/mongo/db/local_catalog/BUILD.bazel

@@ -106,6 +106,7 @@ mongo_cc_library(
         "//src/mongo/db/repl:oplog_visibility_manager",
         "//src/mongo/db/repl:optime",
         "//src/mongo/db/repl:repl_coordinator_interface",  # TODO(SERVER-93876): Remove.
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/storage:oplog_truncate_markers",
         "//src/mongo/db/storage:record_store_base",
     ],
@@ -124,6 +125,7 @@ mongo_cc_library(
         ":durable_catalog_entry_metadata",
         "//src/mongo/db:server_base",
         "//src/mongo/db/op_observer:op_observer_util",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/storage:feature_document_util",
         "//src/mongo/db/storage:ident",
         "//src/mongo/db/storage:mdb_catalog",
@@ -646,7 +648,6 @@ mongo_cc_library(
         "//src/mongo/db:vector_clock",
         "//src/mongo/db/collection_crud",
         "//src/mongo/db/commands:server_status_core",
-        "//src/mongo/db/disagg_storage:server_parameters",
         "//src/mongo/db/index:index_access_method",
         "//src/mongo/db/index:preallocated_container_pool",
         "//src/mongo/db/matcher/doc_validation",
@@ -656,6 +657,7 @@ mongo_cc_library(
         "//src/mongo/db/repl:oplog",
         "//src/mongo/db/repl:repl_server_parameters",
         "//src/mongo/db/repl:repl_settings",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/stats:top",
         "//src/mongo/db/storage:mdb_catalog",
         "//src/mongo/db/storage:record_store_base",

src/mongo/db/local_catalog/collection_options.h

@@ -176,6 +176,8 @@ struct CollectionOptions {
     boost::optional<EncryptedFieldConfig> encryptedFieldConfig;
 
     // When 'true', will use the same recordIds across all nodes in the replica set.
+    // When using disaggregated storage, will be enabled implicitly when the collection
+    // is created.
     bool recordIdsReplicated = false;
 };
 

src/mongo/db/local_catalog/database_impl.cpp

@@ -35,7 +35,6 @@
 #include "mongo/bson/bsonobjbuilder.h"
 #include "mongo/db/audit.h"
 #include "mongo/db/basic_types_gen.h"
-#include "mongo/db/disagg_storage/server_parameters_gen.h"
 #include "mongo/db/index_builds/index_build_block.h"
 #include "mongo/db/index_builds/index_builds_common.h"
 #include "mongo/db/local_catalog/catalog_raii.h"
@@ -65,6 +64,7 @@
 #include "mongo/db/record_id.h"
 #include "mongo/db/repl/oplog.h"
 #include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/server_feature_flags_gen.h"
 #include "mongo/db/server_options.h"
 #include "mongo/db/server_parameter.h"
@@ -158,10 +158,9 @@ RecordId acquireCatalogId(
     OperationContext* opCtx,
     const boost::optional<CreateCollCatalogIdentifier>& createCollCatalogIdentifier,
     MDBCatalog* mdbCatalog) {
-    if (disagg::gDisaggregatedStorageEnabled && createCollCatalogIdentifier.has_value()) {
+    auto& rss = rss::ReplicatedStorageService::get(opCtx);
-        // Replicated catalogIds aren't compatible with standard architecture, as a node may create
+    if (rss.getPersistenceProvider().shouldUseReplicatedCatalogIdentifiers() &&
-        // local collection whose catalogId collides with that of a replicated collection created on
+        createCollCatalogIdentifier.has_value()) {
-        // another node.
         return createCollCatalogIdentifier->catalogId;
     }
     return mdbCatalog->reserveCatalogId(opCtx);
@@ -770,8 +769,10 @@ Collection* DatabaseImpl::_createCollection(
     // Additionally, we do not set the recordIdsReplicated:true option on timeseries and
     // clustered collections because in those cases the recordId is the _id, or on capped
     // collections which utilizes a separate mechanism for ensuring uniform recordIds.
-    if (generatedUUID && !nss.isOnInternalDb() && !optionsWithUUID.timeseries &&
+    const bool collectionTypeSupportsReplicatedRecordIds =
-        !optionsWithUUID.clusteredIndex && !optionsWithUUID.capped &&
+        !optionsWithUUID.timeseries && !optionsWithUUID.clusteredIndex && !optionsWithUUID.capped;
+    const auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
+    if (generatedUUID && !nss.isOnInternalDb() && collectionTypeSupportsReplicatedRecordIds &&
         gFeatureFlagRecordIdsReplicated.isEnabledUseLastLTSFCVWhenUninitialized(
             VersionContext::getDecoration(opCtx),
             serverGlobalParams.featureCompatibility.acquireFCVSnapshot()) &&
@@ -781,6 +782,19 @@ Collection* DatabaseImpl::_createCollection(
                     "Collection will use recordIdsReplicated:true.",
                     "oldValue"_attr = optionsWithUUID.recordIdsReplicated);
         optionsWithUUID.recordIdsReplicated = true;
+    } else if (provider.shouldUseReplicatedRecordIds() && nss.isReplicated() &&
+               !nss.isImplicitlyReplicated() && collectionTypeSupportsReplicatedRecordIds) {
+        tassert(10985561,
+                str::stream() << "Replicated record IDs must be enabled with " << provider.name(),
+                gFeatureFlagRecordIdsReplicated.isEnabledUseLatestFCVWhenUninitialized(
+                    VersionContext::getDecoration(opCtx),
+                    serverGlobalParams.featureCompatibility.acquireFCVSnapshot()));
+        LOGV2_DEBUG(10985560,
+                    2,
+                    "Collection will use recordIdsReplicated:true",
+                    "provider"_attr = provider.name(),
+                    "oldValue"_attr = optionsWithUUID.recordIdsReplicated);
+        optionsWithUUID.recordIdsReplicated = true;
     }
 
     uassert(ErrorCodes::CommandNotSupported,

src/mongo/db/local_catalog/durable_catalog.cpp

@@ -39,6 +39,7 @@
 #include "mongo/db/local_catalog/shard_role_api/transaction_resources.h"
 #include "mongo/db/op_observer/op_observer_util.h"
 #include "mongo/db/operation_context.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/storage/feature_document_util.h"
 #include "mongo/db/storage/kv/kv_engine.h"
 #include "mongo/db/storage/mdb_catalog.h"
@@ -331,10 +332,13 @@ Status createIndex(OperationContext* opCtx,
     auto& ru = *shard_role_details::getRecoveryUnit(opCtx);
     auto storageEngine = opCtx->getServiceContext()->getStorageEngine();
     auto kvEngine = storageEngine->getEngine();
+    auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
+
     invariant(collectionOptions.uuid);
 
-    bool replicateLocalCatalogIdentifiers =
+    bool replicateLocalCatalogIdentifiers = shouldReplicateLocalCatalogIdentifers(
-        shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx));
+        rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+        VersionContext::getDecoration(opCtx));
     if (replicateLocalCatalogIdentifiers) {
         // If a previous attempt at creating this index was rolled back, the ident may still be drop
         // pending. Complete that drop before creating the index if so.
@@ -349,6 +353,7 @@ Status createIndex(OperationContext* opCtx,
     }
 
     Status status = kvEngine->createSortedDataInterface(
+        provider,
         ru,
         nss,
         *collectionOptions.uuid,
@@ -466,8 +471,10 @@ Status dropAndRecreateIndexIdentForResume(OperationContext* opCtx,
         return status;
 
     invariant(collectionOptions.uuid);
+    auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
     status =
-        engine->createSortedDataInterface(*shard_role_details::getRecoveryUnit(opCtx),
+        engine->createSortedDataInterface(provider,
+                                          *shard_role_details::getRecoveryUnit(opCtx),
                                           nss,
                                           *collectionOptions.uuid,
                                           ident,

src/mongo/db/local_catalog/local_oplog_info.cpp

@@ -39,6 +39,7 @@
 #include "mongo/db/repl/oplog.h"
 #include "mongo/db/repl/optime.h"
 #include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/storage/record_store.h"
 #include "mongo/db/storage/recovery_unit.h"
 #include "mongo/db/storage/storage_options.h"
@@ -146,6 +147,10 @@ std::vector<OplogSlot> LocalOplogInfo::getNextOpTimes(OperationContext* opCtx, s
     Timestamp ts;
     // Provide a sample to FlowControl after the `oplogInfo.newOpMutex` is released.
     ON_BLOCK_EXIT([opCtx, &ts, count] {
+        auto& rss = rss::ReplicatedStorageService::get(opCtx);
+        if (!rss.getPersistenceProvider().shouldUseOplogWritesForFlowControlSampling())
+            return;
+
         auto flowControl = FlowControl::get(opCtx);
         if (flowControl) {
             flowControl->sample(ts, count);

src/mongo/db/mongod_main.cpp

@@ -154,17 +154,14 @@
 #include "mongo/db/repl/repl_settings.h"
 #include "mongo/db/repl/replication_consistency_markers_impl.h"
 #include "mongo/db/repl/replication_coordinator.h"
-#include "mongo/db/repl/replication_coordinator_external_state_impl.h"
-#include "mongo/db/repl/replication_coordinator_impl.h"
 #include "mongo/db/repl/replication_coordinator_impl_gen.h"
 #include "mongo/db/repl/replication_process.h"
 #include "mongo/db/repl/replication_recovery.h"
-#include "mongo/db/repl/storage_interface.h"
 #include "mongo/db/repl/storage_interface_impl.h"
-#include "mongo/db/repl/topology_coordinator.h"
 #include "mongo/db/repl/wait_for_majority_service.h"
 #include "mongo/db/replication_state_transition_lock_guard.h"
 #include "mongo/db/request_execution_context.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/s/migration_blocking_operation/multi_update_coordinator.h"
 #include "mongo/db/s/migration_chunk_cloner_source_op_observer.h"
 #include "mongo/db/s/query_analysis_op_observer_configsvr.h"
@@ -606,9 +603,9 @@ ExitCode _initAndListen(ServiceContext* serviceContext) {
         ec != ExitCode::clean)
         return ec;
 
-    FlowControl::set(serviceContext,
+    auto& rss = rss::ReplicatedStorageService::get(serviceContext);
-                     std::make_unique<FlowControl>(
+    auto& serviceLifecycle = rss.getServiceLifecycle();
-                         serviceContext, repl::ReplicationCoordinator::get(serviceContext)));
+    serviceLifecycle.initializeFlowControl(serviceContext);
 
     // If a crash occurred during file-copy based initial sync, we may need to finish or clean up.
     {
@@ -620,8 +617,20 @@ ExitCode _initAndListen(ServiceContext* serviceContext) {
 
     admission::initializeExecutionControl(serviceContext);
 
-    auto lastShutdownState = catalog::startUpStorageEngineAndCollectionCatalog(
+    serviceLifecycle.initializeStorageEngineExtensions(serviceContext);
-        serviceContext, &cc(), StorageEngineInitFlags{}, &startupTimeElapsedBuilder);
+
+    auto lastShutdownState = [&]() {
+        if (rss.getPersistenceProvider().shouldDelayDataAccessDuringStartup()) {
+            // If data isn't ready yet, we shouldn't try to read it.
+            auto initializeStorageEngineOpCtx = serviceContext->makeOperationContext(&cc());
+            return catalog::startUpStorageEngine(initializeStorageEngineOpCtx.get(),
+                                                 StorageEngineInitFlags{},
+                                                 &startupTimeElapsedBuilder);
+        } else {
+            return catalog::startUpStorageEngineAndCollectionCatalog(
+                serviceContext, &cc(), StorageEngineInitFlags{}, &startupTimeElapsedBuilder);
+        }
+    }();
     StorageControl::startStorageControls(serviceContext);
 
     auto logStartupStats = std::make_unique<ScopeGuard<std::function<void()>>>([&] {
@@ -898,7 +907,8 @@ ExitCode _initAndListen(ServiceContext* serviceContext) {
                                        &startupTimeElapsedBuilder);
         replCoord->startup(startupOpCtx.get(), lastShutdownState);
     } else {
-        if (storageEngine->supportsCappedCollections()) {
+        if (rss.getPersistenceProvider().supportsLocalCollections() &&
+            storageEngine->supportsCappedCollections()) {
             logStartup(startupOpCtx.get());
         }
 
@@ -1368,30 +1378,16 @@ auto makeReplicaSetNodeExecutor(ServiceContext* serviceContext) {
             "ReplNodeDbWorkerNetwork", nullptr, makeShardingEgressHooksList(serviceContext)));
 }
 
-auto makeReplicationExecutor(ServiceContext* serviceContext) {
-    ThreadPool::Options tpOptions;
-    tpOptions.threadNamePrefix = "ReplCoord-";
-    tpOptions.poolName = "ReplCoordThreadPool";
-    tpOptions.maxThreads = 50;
-    tpOptions.onCreateThread = [serviceContext](const std::string& threadName) {
-        Client::initThread(threadName,
-                           serviceContext->getService(ClusterRole::ShardServer),
-                           Client::noSession(),
-                           ClientOperationKillableByStepdown{false});
-    };
-    auto hookList = std::make_unique<rpc::EgressMetadataHookList>();
-    hookList->addHook(std::make_unique<rpc::VectorClockMetadataHook>(serviceContext));
-    return executor::ThreadPoolTaskExecutor::create(
-        std::make_unique<ThreadPool>(tpOptions),
-        executor::makeNetworkInterface("ReplNetwork", nullptr, std::move(hookList)));
-}
-
 void setUpReplicaSetDDLHooks(ServiceContext* serviceContext) {
     ReplicaSetDDLTracker::create(serviceContext);
     DirectConnectionDDLHook::create(serviceContext);
 }
 
 void setUpReplication(ServiceContext* serviceContext) {
+    auto& serviceLifecycle =
+        rss::ReplicatedStorageService::get(serviceContext).getServiceLifecycle();
+    serviceLifecycle.initializeStateRequiredForStorageAccess(serviceContext);
+
     repl::StorageInterface::set(serviceContext, std::make_unique<repl::StorageInterfaceImpl>());
     auto storageInterface = repl::StorageInterface::get(serviceContext);
 
@@ -1403,22 +1399,10 @@ void setUpReplication(ServiceContext* serviceContext) {
         serviceContext,
         std::make_unique<repl::ReplicationProcess>(
             storageInterface, std::move(consistencyMarkers), std::move(recovery)));
-    auto replicationProcess = repl::ReplicationProcess::get(serviceContext);
 
-    repl::TopologyCoordinator::Options topoCoordOptions;
+    std::unique_ptr<repl::ReplicationCoordinator> replCoord =
-    topoCoordOptions.maxSyncSourceLagSecs = Seconds(repl::maxSyncSourceLagSecs);
+        serviceLifecycle.initializeReplicationCoordinator(serviceContext);
-    topoCoordOptions.clusterRole = serverGlobalParams.clusterRole;
 
-    auto replCoord = std::make_unique<repl::ReplicationCoordinatorImpl>(
-        serviceContext,
-        getGlobalReplSettings(),
-        std::make_unique<repl::ReplicationCoordinatorExternalStateImpl>(
-            serviceContext, storageInterface, replicationProcess),
-        makeReplicationExecutor(serviceContext),
-        std::make_unique<repl::TopologyCoordinator>(topoCoordOptions),
-        replicationProcess,
-        storageInterface,
-        SecureRandom().nextInt64());
     // Only create a ReplicaSetNodeExecutor if sharding is disabled and replication is enabled.
     // Note that sharding sets up its own executors for scheduling work to remote nodes.
     if (serverGlobalParams.clusterRole.has(ClusterRole::None) &&
@@ -1840,8 +1824,12 @@ void shutdownTask(const ShutdownTaskArgs& shutdownArgs) {
             SectionScopedTimer scopedTimer(serviceContext->getFastClockSource(),
                                            TimedSectionId::killAllOperations,
                                            &shutdownTimeElapsedBuilder);
-            serviceContext->setKillAllOperations(
+            auto& serviceLifecycle =
-                [](const StringData t) { return t == kFTDCThreadName; });
+                rss::ReplicatedStorageService::get(serviceContext).getServiceLifecycle();
+            serviceContext->setKillAllOperations([&serviceLifecycle](const StringData t) {
+                return t == kFTDCThreadName ||
+                    serviceLifecycle.shouldKeepThreadAliveUntilStorageEngineHasShutDown(t);
+            });
 
             if (MONGO_unlikely(pauseWhileKillingOperationsAtShutdown.shouldFail())) {
                 LOGV2_OPTIONS(4701700,
@@ -1987,6 +1975,13 @@ void shutdownTask(const ShutdownTaskArgs& shutdownArgs) {
                                                                         true /* memLeakAllowed */);
     }
 
+    // Depending on the underlying implementation, there may be some state that needs to be shut
+    // down after the replication subsystem and the storage engine.
+    auto& serviceLifecycle =
+        rss::ReplicatedStorageService::get(serviceContext).getServiceLifecycle();
+    serviceLifecycle.shutdownStateRequiredForStorageAccess(serviceContext,
+                                                           &shutdownTimeElapsedBuilder);
+
     // We drop the scope cache because leak sanitizer can't see across the
     // thread we use for proxying MozJS requests. Dropping the cache cleans up
     // the memory and makes leak sanitizer happy.

src/mongo/db/op_msg_fuzzer_router_fixture_test.cpp

@@ -27,7 +27,6 @@
  *    it in the license file.
  */
 
-
 #include "mongo/db/op_msg_fuzzer_router_fixture.h"
 
 #include "mongo/base/string_data.h"

src/mongo/db/op_observer/BUILD.bazel

@@ -141,8 +141,8 @@ mongo_cc_library(
         "//src/mongo:base",
         "//src/mongo/bson/dotted_path:dotted_path_support",  # TODO(SERVER-93876): Remove.
         "//src/mongo/db:shard_role_api",
-        "//src/mongo/db/disagg_storage:server_parameters",
         "//src/mongo/db/local_catalog:collection_options",
+        "//src/mongo/db/rss:replicated_storage_service",
     ],
 )
 

src/mongo/db/op_observer/op_observer_impl.cpp

@@ -66,6 +66,7 @@
 #include "mongo/db/repl/oplog_entry_gen.h"
 #include "mongo/db/repl/read_concern_args.h"
 #include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/server_options.h"
 #include "mongo/db/session/logical_session_id_helpers.h"
 #include "mongo/db/session/session_txn_record_gen.h"
@@ -330,8 +331,9 @@ void OpObserverImpl::onCreateIndex(OperationContext* opCtx,
         return;
     }
 
-    bool replicateLocalCatalogIdentifiers =
+    bool replicateLocalCatalogIdentifiers = shouldReplicateLocalCatalogIdentifers(
-        shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx));
+        rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+        VersionContext::getDecoration(opCtx));
 
     BSONObjBuilder builder;
     // Note that despite using this constant, we are not building a CreateIndexCommand here
@@ -417,7 +419,9 @@ void OpObserverImpl::onStartIndexBuild(OperationContext* opCtx,
     oplogEntry.setNss(nss.getCommandNS());
     oplogEntry.setUuid(collUUID);
     oplogEntry.setObject(oplogEntryBuilder.done());
-    if (shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx))) {
+    if (shouldReplicateLocalCatalogIdentifers(
+            rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+            VersionContext::getDecoration(opCtx))) {
         // TODO (SERVER-109824): Move 'directoryPerDB' and 'directoryForIndexes' to the function
         // parameters.
         oplogEntry.setObject2(BSON("indexes" << o2IndexesArr.arr() << "directoryPerDB"
@@ -1190,7 +1194,9 @@ void OpObserverImpl::onCreateCollection(
     oplogEntry.setNss(collectionName.getCommandNS());
     oplogEntry.setUuid(options.uuid);
     oplogEntry.setObject(MutableOplogEntry::makeCreateCollObject(collectionName, options, idIndex));
-    if (shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx))) {
+    if (shouldReplicateLocalCatalogIdentifers(
+            rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+            VersionContext::getDecoration(opCtx))) {
         invariant(createCollCatalogIdentifier.has_value(),
                   "Missing catalog identifier required to log replicated "
                   "collection");

src/mongo/db/op_observer/op_observer_util.cpp

@@ -33,8 +33,8 @@
 #include "mongo/bson/bsonelement.h"
 #include "mongo/bson/bsonobjbuilder.h"
 #include "mongo/bson/dotted_path/dotted_path_support.h"
-#include "mongo/db/disagg_storage/server_parameters_gen.h"
 #include "mongo/db/global_catalog/shard_key_pattern.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/storage/storage_parameters_gen.h"
 #include "mongo/util/duration.h"
 #include "mongo/util/fail_point.h"
@@ -53,10 +53,9 @@ const OpStateAccumulator::Decoration<std::unique_ptr<ShardingWriteRouter>>
 MONGO_FAIL_POINT_DEFINE(addDestinedRecipient);
 MONGO_FAIL_POINT_DEFINE(sleepBetweenInsertOpTimeGenerationAndLogOp);
 
-bool shouldReplicateLocalCatalogIdentifers(const VersionContext& vCtx) {
+bool shouldReplicateLocalCatalogIdentifers(const rss::PersistenceProvider& provider,
-    if (disagg::gDisaggregatedStorageEnabled) {
+                                           const VersionContext& vCtx) {
-        // Disaggregated storage relies on consistent catalog storage. Safe-guard if FCV is not yet
+    if (provider.shouldUseReplicatedCatalogIdentifiers()) {
-        // initialized despite the feature being enabled.
         return true;
     }
     const auto fcvSnapshot = serverGlobalParams.featureCompatibility.acquireFCVSnapshot();

src/mongo/db/op_observer/op_observer_util.h

@@ -35,6 +35,7 @@
 #include "mongo/db/local_catalog/collection_options.h"
 #include "mongo/db/namespace_string.h"
 #include "mongo/db/op_observer/op_observer.h"
+#include "mongo/db/rss/persistence_provider.h"
 #include "mongo/util/assert_util.h"
 #include "mongo/util/decorable.h"
 #include "mongo/util/fail_point.h"
@@ -54,7 +55,8 @@ extern FailPoint sleepBetweenInsertOpTimeGenerationAndLogOp;
 /**
  * Returns true when local catalog identifiers should be replicated through the oplog.
  */
-bool shouldReplicateLocalCatalogIdentifers(const VersionContext& vCtx);
+bool shouldReplicateLocalCatalogIdentifers(const rss::PersistenceProvider&,
+                                           const VersionContext& vCtx);
 
 /**
  * Returns true if gFeatureFlagPrimaryDrivenIndexBuilds is enabled.

src/mongo/db/repl/BUILD.bazel

@@ -972,6 +972,7 @@ mongo_cc_library(
         "//src/mongo/db/pipeline:change_stream_preimage",
         "//src/mongo/db/query/write_ops",
         "//src/mongo/db/repl/dbcheck",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/session:session_catalog_mongod",
         "//src/mongo/db/stats:counters",
         "//src/mongo/db/stats:server_read_concern_write_concern_metrics",

src/mongo/db/repl/oplog.cpp

@@ -105,6 +105,7 @@
 #include "mongo/db/repl/replication_coordinator.h"
 #include "mongo/db/repl/timestamp_block.h"
 #include "mongo/db/repl/transaction_oplog_application.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/service_context.h"
 #include "mongo/db/session/logical_session_id_gen.h"
 #include "mongo/db/sharding_environment/shard_id.h"
@@ -211,7 +212,10 @@ StringData getInvalidatingReason(const OplogApplication::Mode mode, const bool i
 boost::optional<CreateCollCatalogIdentifier> extractReplicatedCatalogIdentifier(
     OperationContext* opCtx, const OplogEntry& oplogEntry) {
     auto& o2 = oplogEntry.getObject2();
-    if (!o2 || !shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx))) {
+    if (!o2 ||
+        !shouldReplicateLocalCatalogIdentifers(
+            rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+            VersionContext::getDecoration(opCtx))) {
         // Either no catalog identifier information was provided, or replicated local catalog
         // identifiers are not supported.
         return boost::none;
@@ -365,7 +369,9 @@ void createIndexForApplyOps(OperationContext* opCtx,
     IndexBuildInfo indexBuildInfo = [&] {
         auto storageEngine = opCtx->getServiceContext()->getStorageEngine();
         if (!indexMetadata ||
-            !shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx))) {
+            !shouldReplicateLocalCatalogIdentifers(
+                rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+                VersionContext::getDecoration(opCtx))) {
             return IndexBuildInfo(indexSpec, *storageEngine, indexCollection->ns().dbName());
         }
 
@@ -733,8 +739,13 @@ void createOplog(OperationContext* opCtx,
         uow.commit();
     });
 
-    /* sync here so we don't get any surprising lag later when we try to sync */
+    // We cannot guarantee that we have a stable timestamp at this point, but if the persistence
-    service->getStorageEngine()->flushAllFiles(opCtx, /*callerHoldsReadLock*/ false);
+    // provider supports unstable checkpoints, we can take a checkpoint now to avoid any surprising
+    // lag later when we try to sync.
+    auto& rss = rss::ReplicatedStorageService::get(service);
+    if (rss.getPersistenceProvider().supportsUnstableCheckpoints()) {
+        service->getStorageEngine()->flushAllFiles(opCtx, /*callerHoldsReadLock*/ false);
+    }
 }
 
 void createOplog(OperationContext* opCtx) {
@@ -961,7 +972,9 @@ const StringMap<ApplyOpMetadata> kOpsMap = {
           auto swOplogEntry = IndexBuildOplogEntry::parse(
               opCtx,
               entry,
-              shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(opCtx)));
+              shouldReplicateLocalCatalogIdentifers(
+                  rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider(),
+                  VersionContext::getDecoration(opCtx)));
           if (!swOplogEntry.isOK()) {
               return swOplogEntry.getStatus().withContext(
                   "Error parsing 'startIndexBuild' oplog entry");

src/mongo/db/repl/oplog_applier_impl.cpp

@@ -291,7 +291,7 @@ public:
     }
 
     // Should be called after all oplog entries have been processed to handle the deletes that
-    // were not superceded by a later write.
+    // were not superseded by a later write.
     void handleLatestDeletes(std::function<void(OplogEntry*)> handler) {
         std::for_each(_retryImageWrites.begin(),
                       _retryImageWrites.end(),

src/mongo/db/repl/oplog_applier_impl_test.cpp

@@ -103,7 +103,6 @@
 #include "mongo/db/session/session_txn_record_gen.h"
 #include "mongo/db/sharding_environment/shard_id.h"
 #include "mongo/db/stats/counters.h"
-#include "mongo/db/storage/mdb_catalog.h"
 #include "mongo/db/storage/write_unit_of_work.h"
 #include "mongo/db/tenant_id.h"
 #include "mongo/db/timeseries/timeseries_gen.h"
@@ -141,21 +140,6 @@
 namespace mongo {
 namespace repl {
 namespace {
-CreateCollCatalogIdentifier newCatalogIdentifier(OperationContext* opCtx,
-                                                 const DatabaseName& dbName,
-                                                 bool includeIdIndexIdent) {
-    auto storageEngine = opCtx->getServiceContext()->getStorageEngine();
-    auto mdbCatalog = storageEngine->getMDBCatalog();
-    invariant(mdbCatalog);
-
-    CreateCollCatalogIdentifier catalogIdentifier;
-    catalogIdentifier.catalogId = mdbCatalog->reserveCatalogId(opCtx);
-    catalogIdentifier.ident = storageEngine->generateNewCollectionIdent(dbName);
-    if (includeIdIndexIdent) {
-        catalogIdentifier.idIndexIdent = storageEngine->generateNewIndexIdent(dbName);
-    }
-    return catalogIdentifier;
-}
 
 auto parseFromOplogEntryArray(const BSONObj& obj, int elem) {
     BSONElement tsArray;
@@ -632,45 +616,6 @@ TEST_F(OplogApplierImplTest, CreateCollectionCommand) {
     ASSERT_TRUE(collectionExists(_opCtx.get(), nss));
 }
 
-TEST_F(OplogApplierImplTest, CreateCollectionCommandDisaggBasic) {
-    RAIIServerParameterControllerForTest disaggServer("disaggregatedStorageEnabled", true);
-    RAIIServerParameterControllerForTest replicateLocalCatalogInfoController(
-        "featureFlagReplicateLocalCatalogIdentifiers", true);
-
-    NamespaceString nss = NamespaceString::createNamespaceString_forTest("test.t");
-    auto catalogIdentifier =
-        newCatalogIdentifier(_opCtx.get(), nss.dbName(), true /* includeIdIndexIdent*/);
-
-    auto entry =
-        makeCreateCollectionOplogEntry(nextOpTime(),
-                                       nss,
-                                       CollectionOptions{.uuid = UUID::gen()},
-                                       BSON("v" << 2 << "key" << BSON("_id_" << 1) << "name"
-                                                << "_id_") /* idIndex */,
-                                       catalogIdentifier);
-
-    bool applyCmdCalled = false;
-    _opObserver->onCreateCollectionFn =
-        [&](OperationContext* opCtx,
-            const NamespaceString& collNss,
-            const CollectionOptions&,
-            const BSONObj&,
-            const boost::optional<CreateCollCatalogIdentifier>& collCatalogIdentifier) {
-            applyCmdCalled = true;
-            ASSERT_TRUE(opCtx);
-            ASSERT_TRUE(
-                shard_role_details::getLocker(opCtx)->isDbLockedForMode(nss.dbName(), MODE_IX));
-            ASSERT_EQUALS(nss, collNss);
-            ASSERT(collCatalogIdentifier);
-            ASSERT_EQUALS(catalogIdentifier.catalogId, collCatalogIdentifier->catalogId);
-            ASSERT_EQUALS(catalogIdentifier.ident, collCatalogIdentifier->ident);
-        };
-    ASSERT_OK(_applyOplogEntryOrGroupedInsertsWrapper(
-        _opCtx.get(), ApplierOperation{&entry}, OplogApplication::Mode::kInitialSync));
-    ASSERT_TRUE(applyCmdCalled);
-    ASSERT_TRUE(collectionExists(_opCtx.get(), nss));
-}
-
 TEST_F(OplogApplierImplTest, CreateCollectionCommandMultitenant) {
     setServerParameter("multitenancySupport", true);
     setServerParameter("featureFlagRequireTenantID", true);

src/mongo/db/repl/oplog_applier_impl_test_fixture.cpp

@@ -62,6 +62,7 @@
 #include "mongo/db/repl/storage_interface_impl.h"
 #include "mongo/db/session/session_catalog_mongod.h"
 #include "mongo/db/sharding_environment/shard_id.h"
+#include "mongo/db/storage/mdb_catalog.h"
 #include "mongo/db/storage/write_unit_of_work.h"
 #include "mongo/db/tenant_id.h"
 #include "mongo/db/transaction/session_catalog_mongod_transaction_interface_impl.h"
@@ -198,14 +199,18 @@ void OplogApplierImplOpObserver::onCollMod(OperationContext* opCtx,
     onCollModFn(opCtx, nss, uuid, collModCmd, oldCollOptions, indexInfo);
 }
 
+std::unique_ptr<ReplicationCoordinator> OplogApplierImplTest::makeReplCoord(
+    ServiceContext* serviceContext) {
+    return std::make_unique<ReplicationCoordinatorMock>(serviceContext);
+}
+
 void OplogApplierImplTest::setUp() {
     ServiceContextMongoDTest::setUp();
 
     serviceContext = getServiceContext();
     _opCtx = cc().makeOperationContext();
 
-    ReplicationCoordinator::set(serviceContext,
+    ReplicationCoordinator::set(serviceContext, makeReplCoord(serviceContext));
-                                std::make_unique<ReplicationCoordinatorMock>(serviceContext));
     ASSERT_OK(ReplicationCoordinator::get(_opCtx.get())->setFollowerMode(MemberState::RS_PRIMARY));
 
     StorageInterface::set(serviceContext, std::make_unique<StorageInterfaceImpl>());
@@ -625,5 +630,21 @@ void createIndex(OperationContext* opCtx,
         opCtx, collUUID, spec, IndexBuildsManager::IndexConstraints::kEnforce, false);
 }
 
+CreateCollCatalogIdentifier newCatalogIdentifier(OperationContext* opCtx,
+                                                 const DatabaseName& dbName,
+                                                 bool includeIdIndexIdent) {
+    auto storageEngine = opCtx->getServiceContext()->getStorageEngine();
+    auto mdbCatalog = storageEngine->getMDBCatalog();
+    invariant(mdbCatalog);
+
+    CreateCollCatalogIdentifier catalogIdentifier;
+    catalogIdentifier.catalogId = mdbCatalog->reserveCatalogId(opCtx);
+    catalogIdentifier.ident = storageEngine->generateNewCollectionIdent(dbName);
+    if (includeIdIndexIdent) {
+        catalogIdentifier.idIndexIdent = storageEngine->generateNewIndexIdent(dbName);
+    }
+    return catalogIdentifier;
+}
+
 }  // namespace repl
 }  // namespace mongo

src/mongo/db/repl/oplog_applier_impl_test_fixture.h

@@ -277,6 +277,7 @@ protected:
         return OpTime(Timestamp(Seconds(lastSecond++), 0), 1LL);
     }
 
+    virtual std::unique_ptr<ReplicationCoordinator> makeReplCoord(ServiceContext*);
     void setUp() override;
     void tearDown() override;
 
@@ -415,6 +416,13 @@ void createIndex(OperationContext* opCtx,
                  UUID collUUID,
                  const BSONObj& spec);
 
-}  // namespace MONGO_MOD_PUB repl
+/**
+ * Generate a new catalog identifier.
+ */
+CreateCollCatalogIdentifier newCatalogIdentifier(OperationContext* opCtx,
+                                                 const DatabaseName& dbName,
+                                                 bool includeIdIndexIdent);
 
+
+}  // namespace MONGO_MOD_PUB repl
 }  // namespace mongo

src/mongo/db/repl/optime.h

@@ -212,4 +212,10 @@ public:
 };
 
 std::ostream& operator<<(std::ostream& out, const OpTimeAndWallTime& opTime);
+
+// A convenience class for holding both a Timestamp and a Date_t.
+struct TimestampAndWallTime {
+    Timestamp timestamp;
+    Date_t wallTime;
+};
 }  // namespace mongo::repl

src/mongo/db/repl/rollback_impl_test.cpp

@@ -59,6 +59,7 @@
 #include "mongo/db/repl/replication_consistency_markers.h"
 #include "mongo/db/repl/replication_process.h"
 #include "mongo/db/repl/rollback_test_fixture.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/server_options.h"
 #include "mongo/db/service_context.h"
 #include "mongo/db/session/logical_session_id.h"
@@ -2250,7 +2251,9 @@ TEST_F(RollbackImplObserverInfoTest,
     auto uuid = UUID::gen();
 
     BSONObj indexObj;
-    if (shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(_opCtx.get()))) {
+    if (shouldReplicateLocalCatalogIdentifers(
+            rss::ReplicatedStorageService::get(_opCtx.get()).getPersistenceProvider(),
+            VersionContext::getDecoration(_opCtx.get()))) {
         indexObj = BSON("createIndexes" << nss.coll() << "spec"
                                         << BSON("v" << 2 << "key"
                                                     << "x"

src/mongo/db/repl/storage_timestamp_test.cpp

@@ -106,6 +106,7 @@
 #include "mongo/db/repl/storage_interface.h"
 #include "mongo/db/repl/storage_interface_impl.h"
 #include "mongo/db/repl/timestamp_block.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/service_context.h"
 #include "mongo/db/service_context_d_test_fixture.h"
 #include "mongo/db/session/logical_session_id.h"
@@ -916,7 +917,9 @@ public:
     }
 
     StringData indexNameOplogField() const {
-        return shouldReplicateLocalCatalogIdentifers(VersionContext::getDecoration(_opCtx))
+        return shouldReplicateLocalCatalogIdentifers(
+                   rss::ReplicatedStorageService::get(_opCtx).getPersistenceProvider(),
+                   VersionContext::getDecoration(_opCtx))
             ? "o.spec.name"
             : "o.name";
     }
@@ -3088,14 +3091,16 @@ TEST_F(StorageTimestampTest, CreateCollectionWithSystemIndex) {
     // supports 2 phase index build.
     indexStartTs = op.getTimestamp();
     indexCreateTs =
-        repl::OplogEntry(queryOplog(BSON("op" << "c"
+        repl::OplogEntry(
-                                              << "ns" << nss.getCommandNS().ns_forTest()
+            queryOplog(BSON(
-                                              << "o.createIndexes" << nss.coll()
+                "op" << "c"
-                                              << (shouldReplicateLocalCatalogIdentifers(
+                     << "ns" << nss.getCommandNS().ns_forTest() << "o.createIndexes" << nss.coll()
-                                                      VersionContext::getDecoration(_opCtx))
+                     << (shouldReplicateLocalCatalogIdentifers(
-                                                      ? "o.spec.name"
+                             rss::ReplicatedStorageService::get(_opCtx).getPersistenceProvider(),
-                                                      : "o.name")
+                             VersionContext::getDecoration(_opCtx))
-                                              << "user_1_db_1")))
+                             ? "o.spec.name"
+                             : "o.name")
+                     << "user_1_db_1")))
             .getTimestamp();
     indexCompleteTs = indexCreateTs;
 

src/mongo/db/rss/BUILD.bazel

@@ -0,0 +1,77 @@
+load("//bazel:mongo_src_rules.bzl", "mongo_cc_benchmark", "mongo_cc_library", "mongo_cc_unit_test")
+
+package(default_visibility = ["//visibility:public"])
+
+exports_files(
+    glob([
+        "*.h",
+        "*.cpp",
+    ]),
+)
+
+mongo_cc_library(
+    name = "replicated_storage_service",
+    srcs = [
+        "replicated_storage_service.cpp",
+    ],
+    hdrs = [
+        "replicated_storage_service.h",
+    ],
+    deps = [
+        "//src/mongo:base",
+        "//src/mongo/db:service_context",
+        "//src/mongo/db/rss:persistence_provider",
+        "//src/mongo/db/rss:service_lifecycle",
+    ],
+)
+
+mongo_cc_library(
+    name = "persistence_provider",
+    hdrs = [
+        "persistence_provider.h",
+    ],
+    deps = [
+        "//src/mongo:base",
+        "//src/mongo/db:service_context",
+    ],
+)
+
+mongo_cc_library(
+    name = "service_lifecycle",
+    hdrs = [
+        "service_lifecycle.h",
+    ],
+    deps = [
+        "//src/mongo:base",
+        "//src/mongo/db:service_context",
+        "//src/mongo/db/repl:repl_coordinator_interface",
+    ],
+)
+
+mongo_cc_library(
+    name = "persistence_provider_impl",
+    deps = [
+        "//src/mongo:base",
+        "//src/mongo/db/rss:persistence_provider",
+        "//src/mongo/db/rss/attached_storage:attached_persistence_provider",
+    ] + select({
+        "//bazel/config:build_atlas_enabled": [
+            "//src/mongo/db/modules/atlas/src/disagg_storage:disaggregated_persistence_provider",
+        ],
+        "//conditions:default": [],
+    }),
+)
+
+mongo_cc_library(
+    name = "service_lifecycle_impl",
+    deps = [
+        "//src/mongo:base",
+        "//src/mongo/db/rss:service_lifecycle",
+        "//src/mongo/db/rss/attached_storage:attached_service_lifecycle",
+    ] + select({
+        "//bazel/config:build_atlas_enabled": [
+            "//src/mongo/db/modules/atlas/src/disagg_storage:disaggregated_service_lifecycle",
+        ],
+        "//conditions:default": [],
+    }),
+)

src/mongo/db/rss/OWNERS.yml

@@ -0,0 +1,6 @@
+version: 1.0.0
+filters:
+  - "*":
+    approvers:
+      - 10gen/server-storage-execution
+      - 10gen/server-replication

src/mongo/db/rss/attached_storage/BUILD.bazel

@@ -0,0 +1,43 @@
+load("//bazel:mongo_src_rules.bzl", "mongo_cc_benchmark", "mongo_cc_library", "mongo_cc_unit_test")
+
+package(default_visibility = ["//visibility:public"])
+
+exports_files(
+    glob([
+        "*.h",
+        "*.cpp",
+    ]),
+)
+
+mongo_cc_library(
+    name = "attached_persistence_provider",
+    srcs = [
+        "attached_persistence_provider.cpp",
+    ],
+    hdrs = [
+        "attached_persistence_provider.h",
+    ],
+    deps = [
+        "//src/mongo:base",
+        "//src/mongo/db:service_context",
+        "//src/mongo/db/rss:replicated_storage_service",
+    ],
+)
+
+mongo_cc_library(
+    name = "attached_service_lifecycle",
+    srcs = [
+        "attached_service_lifecycle.cpp",
+    ],
+    hdrs = [
+        "attached_service_lifecycle.h",
+    ],
+    deps = [
+        "//src/mongo:base",
+        "//src/mongo/db:service_context",
+        "//src/mongo/db/admission:flow_control",
+        "//src/mongo/db/repl:repl_coordinator_impl",
+        "//src/mongo/db/repl:serveronly_repl",
+        "//src/mongo/db/rss:replicated_storage_service",
+    ],
+)

src/mongo/db/rss/attached_storage/attached_persistence_provider.cpp

@@ -0,0 +1,101 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#include "mongo/db/rss/attached_storage/attached_persistence_provider.h"
+
+#include "mongo/base/string_data.h"
+#include "mongo/db/rss/replicated_storage_service.h"
+#include "mongo/db/service_context.h"
+
+namespace mongo::rss {
+namespace {
+// Checkpoint every 60 seconds by default.
+constexpr double kDefaultAttachedSyncDelaySeconds = 60.0;
+
+ServiceContext::ConstructorActionRegisterer registerAttachedPersistenceProvider{
+    "AttachedPersistenceProvider", [](ServiceContext* service) {
+        auto& rss = ReplicatedStorageService::get(service);
+        rss.setPersistenceProvider(std::make_unique<AttachedPersistenceProvider>());
+    }};
+
+}  // namespace
+
+std::string AttachedPersistenceProvider::name() const {
+    return "Attached Storage";
+}
+
+boost::optional<Timestamp> AttachedPersistenceProvider::getSentinelDataTimestamp() const {
+    return boost::none;
+}
+
+std::string AttachedPersistenceProvider::getWiredTigerConfig(int) const {
+    return "";
+}
+
+bool AttachedPersistenceProvider::shouldUseReplicatedCatalogIdentifiers() const {
+    return false;
+}
+
+bool AttachedPersistenceProvider::shouldUseReplicatedRecordIds() const {
+    return false;
+}
+
+bool AttachedPersistenceProvider::shouldUseOplogWritesForFlowControlSampling() const {
+    return true;
+}
+
+bool AttachedPersistenceProvider::shouldStepDownForShutdown() const {
+    return true;
+}
+
+bool AttachedPersistenceProvider::shouldDelayDataAccessDuringStartup() const {
+    return false;
+}
+
+bool AttachedPersistenceProvider::shouldAvoidDuplicateCheckpoints() const {
+    return false;
+}
+
+bool AttachedPersistenceProvider::supportsLocalCollections() const {
+    return true;
+}
+
+bool AttachedPersistenceProvider::supportsUnstableCheckpoints() const {
+    return true;
+}
+
+bool AttachedPersistenceProvider::supportsTableLogging() const {
+    return true;
+}
+
+bool AttachedPersistenceProvider::supportsMultiDocumentTransactions() const {
+    return true;
+}
+
+}  // namespace mongo::rss

src/mongo/db/rss/attached_storage/attached_persistence_provider.h

@@ -0,0 +1,105 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#pragma once
+
+#include "mongo/db/rss/persistence_provider.h"
+
+namespace mongo::rss {
+
+class AttachedPersistenceProvider : public PersistenceProvider {
+public:
+    std::string name() const override;
+
+    /**
+     * We do not have any specific initialization requirements.
+     */
+    boost::optional<Timestamp> getSentinelDataTimestamp() const override;
+
+    /**
+     * We do not have any additional WT config to add.
+     */
+    std::string getWiredTigerConfig(int) const override;
+
+    /**
+     * Replicated catalog identifiers aren't compatible with attached storage as of right now, as a
+     * node may create a local collection whose catalog identifier collides with that of a
+     * replicated collection created on another node.
+     */
+    bool shouldUseReplicatedCatalogIdentifiers() const override;
+
+    /**
+     * Attached storage does not require replicated RecordIds to function correctly.
+     */
+    bool shouldUseReplicatedRecordIds() const override;
+
+    /**
+     * Flow control is based on the rate of generation of oplog data and the ability of the
+     * secondaries to keep the majority commit point relatively up-to-date.
+     */
+    bool shouldUseOplogWritesForFlowControlSampling() const override;
+
+    /**
+     * Stepping down prior to shut down allows for a graceful and quick election most of the time.
+     */
+    bool shouldStepDownForShutdown() const override;
+
+    /**
+     * We can safely initialize the catalog immediately after starting the storage engine.
+     */
+    bool shouldDelayDataAccessDuringStartup() const override;
+
+    /**
+     * Running a duplicate checkpoint for a given timestamp has little effect other than being
+     * slightly inefficient, so there's no need to use extra synchronization to avoid it.
+     */
+    bool shouldAvoidDuplicateCheckpoints() const override;
+
+    /**
+     * We can support local, fully unreplicated collections.
+     */
+    bool supportsLocalCollections() const override;
+
+    /**
+     * We can support unstable checkpoints.
+     */
+    bool supportsUnstableCheckpoints() const override;
+
+    /**
+     * We can support table logging.
+     */
+    bool supportsTableLogging() const override;
+
+    /**
+     * We can support multi-document transactions.
+     */
+    bool supportsMultiDocumentTransactions() const override;
+};
+
+}  // namespace mongo::rss

src/mongo/db/rss/attached_storage/attached_service_lifecycle.cpp

@@ -0,0 +1,126 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#include "mongo/db/rss/attached_storage/attached_service_lifecycle.h"
+
+#include "mongo/db/admission/flow_control.h"
+#include "mongo/db/global_settings.h"
+#include "mongo/db/repl/replication_consistency_markers_impl.h"
+#include "mongo/db/repl/replication_coordinator_external_state_impl.h"
+#include "mongo/db/repl/replication_coordinator_impl.h"
+#include "mongo/db/repl/storage_interface.h"
+#include "mongo/db/repl/topology_coordinator.h"
+#include "mongo/db/rss/replicated_storage_service.h"
+#include "mongo/db/storage/storage_options.h"
+#include "mongo/executor/network_interface_factory.h"
+#include "mongo/executor/thread_pool_task_executor.h"
+#include "mongo/rpc/metadata/egress_metadata_hook_list.h"
+
+namespace mongo::rss {
+namespace {
+// Checkpoint every 60 seconds by default.
+constexpr double kDefaultAttachedSyncDelaySeconds = 60.0;
+
+ServiceContext::ConstructorActionRegisterer registerAttachedServiceLifecycle{
+    "AttachedServiceLifecycle", [](ServiceContext* service) {
+        auto& rss = ReplicatedStorageService::get(service);
+        rss.setServiceLifecycle(std::make_unique<AttachedServiceLifecycle>());
+    }};
+
+auto makeReplicationExecutor(ServiceContext* serviceContext) {
+    ThreadPool::Options tpOptions;
+    tpOptions.threadNamePrefix = "ReplCoord-";
+    tpOptions.poolName = "ReplCoordThreadPool";
+    tpOptions.maxThreads = 50;
+    tpOptions.onCreateThread = [serviceContext](const std::string& threadName) {
+        Client::initThread(threadName,
+                           serviceContext->getService(ClusterRole::ShardServer),
+                           Client::noSession(),
+                           ClientOperationKillableByStepdown{false});
+    };
+    auto hookList = std::make_unique<rpc::EgressMetadataHookList>();
+    hookList->addHook(std::make_unique<rpc::VectorClockMetadataHook>(serviceContext));
+    return executor::ThreadPoolTaskExecutor::create(
+        std::make_unique<ThreadPool>(tpOptions),
+        executor::makeNetworkInterface("ReplNetwork", nullptr, std::move(hookList)));
+}
+}  // namespace
+
+AttachedServiceLifecycle::AttachedServiceLifecycle()
+    : _initializedUsingDefaultSyncDelay{[]() {
+          if (storageGlobalParams.syncdelay.load() < 0.0) {
+              storageGlobalParams.syncdelay.store(kDefaultAttachedSyncDelaySeconds);
+              return true;
+          }  // namespace mongo::rss
+          return false;
+      }()} {}
+
+void AttachedServiceLifecycle::initializeFlowControl(ServiceContext* svcCtx) {
+    FlowControl::set(
+        svcCtx, std::make_unique<FlowControl>(svcCtx, repl::ReplicationCoordinator::get(svcCtx)));
+}
+
+void AttachedServiceLifecycle::initializeStorageEngineExtensions(ServiceContext*) {}
+
+std::unique_ptr<repl::ReplicationCoordinator>
+AttachedServiceLifecycle::initializeReplicationCoordinator(ServiceContext* svcCtx) {
+    auto storageInterface = repl::StorageInterface::get(svcCtx);
+    auto replicationProcess = repl::ReplicationProcess::get(svcCtx);
+
+    repl::TopologyCoordinator::Options topoCoordOptions;
+    topoCoordOptions.maxSyncSourceLagSecs = Seconds(repl::maxSyncSourceLagSecs);
+    topoCoordOptions.clusterRole = serverGlobalParams.clusterRole;
+
+    return std::make_unique<repl::ReplicationCoordinatorImpl>(
+        svcCtx,
+        getGlobalReplSettings(),
+        std::make_unique<repl::ReplicationCoordinatorExternalStateImpl>(
+            svcCtx, storageInterface, replicationProcess),
+        makeReplicationExecutor(svcCtx),
+        std::make_unique<repl::TopologyCoordinator>(topoCoordOptions),
+        replicationProcess,
+        storageInterface,
+        SecureRandom().nextInt64());
+}
+
+void AttachedServiceLifecycle::initializeStateRequiredForStorageAccess(ServiceContext*) {}
+
+void AttachedServiceLifecycle::shutdownStateRequiredForStorageAccess(ServiceContext*,
+                                                                     BSONObjBuilder*) {}
+
+bool AttachedServiceLifecycle::initializedUsingDefaultSyncDelay() const {
+    return _initializedUsingDefaultSyncDelay;
+}
+
+bool AttachedServiceLifecycle::shouldKeepThreadAliveUntilStorageEngineHasShutDown(
+    const StringData) const {
+    return false;
+}
+
+}  // namespace mongo::rss

src/mongo/db/rss/attached_storage/attached_service_lifecycle.h

@@ -0,0 +1,77 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#pragma once
+
+#include "mongo/db/rss/service_lifecycle.h"
+
+namespace mongo::rss {
+
+class AttachedServiceLifecycle : public ServiceLifecycle {
+public:
+    AttachedServiceLifecycle();
+
+    /**
+     * Initializes flow control based on oplog write rate.
+     */
+    void initializeFlowControl(ServiceContext*) override;
+
+    /**
+     * There are no storage engine extensions utilized.
+     */
+    void initializeStorageEngineExtensions(ServiceContext*) override;
+
+    /**
+     * Initializes a 'repl::ReplicationCoordinatorImpl'.
+     */
+    std::unique_ptr<repl::ReplicationCoordinator> initializeReplicationCoordinator(
+        ServiceContext*) override;
+
+    /**
+     * There is no additional state required for storage access.
+     */
+    void initializeStateRequiredForStorageAccess(ServiceContext*) override;
+
+    /**
+     * There is no additional state required for storage access.
+     */
+    void shutdownStateRequiredForStorageAccess(ServiceContext*, BSONObjBuilder*) override;
+
+    bool initializedUsingDefaultSyncDelay() const override;
+
+    /**
+     * There are no specific persistence threads that must outlive the storage engine.
+     */
+    bool shouldKeepThreadAliveUntilStorageEngineHasShutDown(StringData) const override;
+
+private:
+    const bool _initializedUsingDefaultSyncDelay;
+};
+
+}  // namespace mongo::rss

src/mongo/db/rss/persistence_provider.h

@@ -0,0 +1,131 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#pragma once
+
+#include "mongo/base/string_data.h"
+#include "mongo/bson/bsonelement.h"
+
+#include <string>
+#include <utility>
+
+#include <boost/optional.hpp>
+
+
+namespace mongo {
+namespace rss {
+
+/**
+ * This class provides an abstraction around the persistence layer underlying the storage and
+ * replication subsystems. Depending on the configuration, the implementation may be backed by a
+ * local filesystem, a remote service, etc. The interface is built primarily around capabilities and
+ * expected behaviors, allowing consumers to act based on these flags, rather than needing to reason
+ * about how a particular provider would behave in a given context.
+ */
+class PersistenceProvider {
+public:
+    virtual ~PersistenceProvider() = default;
+
+    /**
+     * The name of this provider, for use in e.g. logging and error messages.
+     */
+    virtual std::string name() const = 0;
+
+    /**
+     * If not none, the KVEngine will use the returned Timestamp during initialization as the
+     * initial data timestamp.
+     */
+    virtual boost::optional<Timestamp> getSentinelDataTimestamp() const = 0;
+
+    /**
+     * Additional configuration that shoudld be added to the WiredTiger config string for the
+     * 'wiredtiger_open' call. The 'flattenLeafPageDelta' is expected to be the corresponding
+     * WiredTigerConfig member value.
+     */
+    virtual std::string getWiredTigerConfig(int flattenLeafPageDelta) const = 0;
+
+    /**
+     * If true, the provider expects that all catalog identifiers will be replicated and identical
+     * between nodes.
+     */
+    virtual bool shouldUseReplicatedCatalogIdentifiers() const = 0;
+
+    /**
+     * If true, the provider expects that RecordIds will be replicated (either explicitly or
+     * implicitly) and identical between nodes.
+     */
+    virtual bool shouldUseReplicatedRecordIds() const = 0;
+
+    /**
+     * If true, writes to the oplog should be used as the unit of progress for flow control
+     * sampling.
+     */
+    virtual bool shouldUseOplogWritesForFlowControlSampling() const = 0;
+
+    /**
+     * If true, the node should step down prior to shutdown in order to minimize unavailability.
+     */
+    virtual bool shouldStepDownForShutdown() const = 0;
+
+    /**
+     * If true, data may not be availabile immediately after starting the storage engine, so systems
+     * like the catalog should not be initialized immediately.
+     */
+    virtual bool shouldDelayDataAccessDuringStartup() const = 0;
+
+    /**
+     * If true, the system should take precautions to avoid taking multiple checkopints for the same
+     * stable timestamp. The underlying key-value engine likely does not provide the necessary
+     * coordination by default.
+     */
+    virtual bool shouldAvoidDuplicateCheckpoints() const = 0;
+
+    /**
+     * If true, the storage provider supports the use of local, unreplicated collections.
+     */
+    virtual bool supportsLocalCollections() const = 0;
+
+    /**
+     * If true, the provider can support unstable checkpoints.
+     */
+    virtual bool supportsUnstableCheckpoints() const = 0;
+
+    /**
+     * If true, the provider can support logging (i.e. journaling) on individual tables.
+     */
+    virtual bool supportsTableLogging() const = 0;
+
+    /**
+     * If true, the provider supports multi-document transactions.
+     */
+    virtual bool supportsMultiDocumentTransactions() const = 0;
+};
+
+}  // namespace rss
+}  // namespace mongo

src/mongo/db/rss/replicated_storage_service.cpp

@@ -0,0 +1,63 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#include "mongo/db/rss/replicated_storage_service.h"
+
+namespace mongo::rss {
+namespace {
+const auto getReplicatedStorageService =
+    ServiceContext::declareDecoration<ReplicatedStorageService>();
+}  // namespace
+
+ReplicatedStorageService& ReplicatedStorageService::get(ServiceContext* svcCtx) {
+    return getReplicatedStorageService(svcCtx);
+}
+
+ReplicatedStorageService& ReplicatedStorageService::get(OperationContext* opCtx) {
+    return get(opCtx->getServiceContext());
+}
+
+PersistenceProvider& ReplicatedStorageService::getPersistenceProvider() {
+    invariant(_provider);
+    return *_provider;
+}
+void ReplicatedStorageService::setPersistenceProvider(std::unique_ptr<PersistenceProvider>&& p) {
+    _provider = std::move(p);
+}
+
+ServiceLifecycle& ReplicatedStorageService::getServiceLifecycle() {
+    invariant(_lifecycle);
+    return *_lifecycle;
+}
+
+void ReplicatedStorageService::setServiceLifecycle(std::unique_ptr<ServiceLifecycle>&& l) {
+    _lifecycle = std::move(l);
+}
+
+}  // namespace mongo::rss

src/mongo/db/rss/replicated_storage_service.h

@@ -0,0 +1,55 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#pragma once
+
+#include "mongo/db/operation_context.h"
+#include "mongo/db/rss/persistence_provider.h"
+#include "mongo/db/rss/service_lifecycle.h"
+#include "mongo/db/service_context.h"
+
+namespace mongo::rss {
+
+class ReplicatedStorageService {
+public:
+    static ReplicatedStorageService& get(ServiceContext*);
+    static ReplicatedStorageService& get(OperationContext*);
+
+    PersistenceProvider& getPersistenceProvider();
+    void setPersistenceProvider(std::unique_ptr<PersistenceProvider>&&);
+
+    ServiceLifecycle& getServiceLifecycle();
+    void setServiceLifecycle(std::unique_ptr<ServiceLifecycle>&&);
+
+private:
+    std::unique_ptr<PersistenceProvider> _provider;
+    std::unique_ptr<ServiceLifecycle> _lifecycle;
+};
+
+}  // namespace mongo::rss

src/mongo/db/rss/service_lifecycle.h

@@ -0,0 +1,97 @@
+/**
+ *    Copyright (C) 2025-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#pragma once
+
+#include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/service_context.h"
+
+#include <memory>
+#include <string>
+#include <utility>
+
+namespace mongo {
+namespace rss {
+
+/**
+ * This class provides an abstraction for a set of functionalities related to the service lifecycle
+ * (i.e. startup and shutdown).
+ *
+ * The implementation details are generally closely related to the configured 'PersistenceProvider',
+ * but we separate it out from that class since that class is primarily focused on
+ * capabilities/behaviors, while this class instead represents a set of setup/teardown and related
+ * routines.
+ */
+class ServiceLifecycle {
+public:
+    virtual ~ServiceLifecycle() = default;
+
+    /**
+     * Initializes the flow control algorithm for the current service configuration.
+     */
+    virtual void initializeFlowControl(ServiceContext*) = 0;
+
+    /**
+     * Initializes any storage engine extensions necessary for the current service configuration.
+     */
+    virtual void initializeStorageEngineExtensions(ServiceContext*) = 0;
+
+    /**
+     * Initializes and returns the replication coordinator appropriate for the current service
+     * configuration.
+     */
+    virtual std::unique_ptr<repl::ReplicationCoordinator> initializeReplicationCoordinator(
+        ServiceContext*) = 0;
+
+    /**
+     * Initializes any state required to access 'repl::StorageInterface'. This method will be run
+     * prior to 'initializeReplicationCoordinator'.
+     */
+    virtual void initializeStateRequiredForStorageAccess(ServiceContext*) = 0;
+
+    /**
+     * Tears down any state set up by 'initializeStateRequiredForStorageAccess'.
+     */
+    virtual void shutdownStateRequiredForStorageAccess(ServiceContext*, BSONObjBuilder*) = 0;
+
+    /**
+     * If true, this instance was initialized using the default syncdelay parameter rather than any
+     * user-configured value.
+     */
+    virtual bool initializedUsingDefaultSyncDelay() const = 0;
+
+    /**
+     * If true, the named thread must be kept alive until the storage engine has shut down.
+     */
+    virtual bool shouldKeepThreadAliveUntilStorageEngineHasShutDown(
+        StringData threadName) const = 0;
+};
+
+}  // namespace rss
+}  // namespace mongo

src/mongo/db/service_context_d_test_fixture.cpp

@@ -216,7 +216,7 @@ MongoDScopedGlobalServiceContextForTest::~MongoDScopedGlobalServiceContextForTes
     std::swap(storageGlobalParams.engineSetByUser, _stashedStorageParams.engineSetByUser);
     std::swap(storageGlobalParams.repair, _stashedStorageParams.repair);
 
-    storageGlobalParams.reset();
+    storageGlobalParams.reset_forTest();
 }
 
 }  // namespace mongo

src/mongo/db/startup_recovery.cpp

@@ -89,6 +89,7 @@
 #include "mongo/db/repl/repl_set_member_in_standalone_mode.h"
 #include "mongo/db/repl/replication_coordinator.h"
 #include "mongo/db/repl/storage_interface.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/server_feature_flags_gen.h"
 #include "mongo/db/service_context.h"
 #include "mongo/db/storage/control/journal_flusher.h"
@@ -878,6 +879,14 @@ void startupRecovery(OperationContext* opCtx,
                      StorageEngine* storageEngine,
                      StorageEngine::LastShutdownState lastShutdownState,
                      BSONObjBuilder* startupTimeElapsedBuilder = nullptr) {
+    auto& rss = rss::ReplicatedStorageService::get(opCtx);
+    if (rss.getPersistenceProvider().shouldDelayDataAccessDuringStartup()) {
+        LOGV2(10985327,
+              "Skip startupRecovery; it will be handled later when WT loads the "
+              "checkpoint");
+        return;
+    }
+
     invariant(!storageGlobalParams.repair);
 
     ServiceContext* svcCtx = opCtx->getServiceContext();

src/mongo/db/storage/BUILD.bazel

@@ -262,6 +262,7 @@ mongo_cc_library(
     name = "mdb_catalog",
     srcs = [
         "mdb_catalog.cpp",
+        "//src/mongo/db/rss:persistence_provider.h",
         "//src/mongo/db/storage/kv:kv_engine.h",
     ],
     hdrs = [
@@ -270,6 +271,7 @@ mongo_cc_library(
     deps = [
         ":record_store_base",
         "//src/mongo/db:server_base",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/storage:feature_document_util",
         "//src/mongo/db/storage:ident",
     ],
@@ -423,6 +425,7 @@ mongo_cc_library(
     ],
     hdrs = [
         "oplog_truncate_markers.h",
+        "//src/mongo/db/rss:persistence_provider.h",
         "//src/mongo/db/storage/kv:kv_engine.h",
     ],
     deps = [
@@ -501,6 +504,7 @@ mongo_cc_library(
         ":storage_repair_observer",
         "//src/mongo/db:server_base",
         "//src/mongo/db:shard_role",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/storage/kv:kv_drop_pending_ident_reaper",
     ],
 )

src/mongo/db/storage/devnull/BUILD.bazel

@@ -18,6 +18,7 @@ mongo_cc_library(
     hdrs = [
         "devnull_kv_engine.h",
         "ephemeral_catalog_record_store.h",
+        "//src/mongo/db/rss:persistence_provider.h",
         "//src/mongo/db/storage/kv:kv_engine.h",
     ],
     deps = [

src/mongo/db/storage/devnull/devnull_kv_engine.h

@@ -68,7 +68,8 @@ public:
 
     std::unique_ptr<RecoveryUnit> newRecoveryUnit() override;
 
-    Status createRecordStore(const NamespaceString& nss,
+    Status createRecordStore(const rss::PersistenceProvider&,
+                             const NamespaceString& nss,
                              StringData ident,
                              const RecordStore::Options& options) override {
         return Status::OK();
@@ -89,6 +90,7 @@ public:
                                                           KeyFormat keyFormat) override;
 
     Status createSortedDataInterface(
+        const rss::PersistenceProvider&,
         RecoveryUnit&,
         const NamespaceString& nss,
         const UUID& uuid,

src/mongo/db/storage/kv/kv_drop_pending_ident_reaper_test.cpp

@@ -90,7 +90,8 @@ public:
         return nullptr;
     }
 
-    Status createRecordStore(const NamespaceString& nss,
+    Status createRecordStore(const rss::PersistenceProvider&,
+                             const NamespaceString& nss,
                              StringData ident,
                              const RecordStore::Options& options) override {
         return Status::OK();
@@ -108,6 +109,7 @@ public:
         return {};
     }
     Status createSortedDataInterface(
+        const rss::PersistenceProvider&,
         RecoveryUnit&,
         const NamespaceString& nss,
         const UUID& uuid,

src/mongo/db/storage/kv/kv_engine.h

@@ -32,6 +32,7 @@
 #include "mongo/base/status.h"
 #include "mongo/base/string_data.h"
 #include "mongo/bson/timestamp.h"
+#include "mongo/db/rss/persistence_provider.h"
 #include "mongo/db/storage/compact_options.h"
 #include "mongo/db/storage/record_store.h"
 #include "mongo/db/storage/sorted_data_interface.h"
@@ -126,7 +127,8 @@ public:
      *
      * Creates a 'RecordStore' and generated from the provided 'options'.
      */
-    virtual Status createRecordStore(const NamespaceString& nss,
+    virtual Status createRecordStore(const rss::PersistenceProvider&,
+                                     const NamespaceString& nss,
                                      StringData ident,
                                      const RecordStore::Options& options) = 0;
 
@@ -201,6 +203,7 @@ public:
     virtual bool underCachePressure(int concurrentWriteOuts, int concurrentReadOuts) = 0;
 
     virtual Status createSortedDataInterface(
+        const rss::PersistenceProvider&,
         RecoveryUnit&,
         const NamespaceString& nss,
         const UUID& uuid,
@@ -257,10 +260,11 @@ public:
      * This recovery process makes no guarantees about the integrity of data recovered or even that
      * it still exists when recovered.
      */
-    virtual Status recoverOrphanedIdent(const NamespaceString& nss,
+    virtual Status recoverOrphanedIdent(const rss::PersistenceProvider& provider,
+                                        const NamespaceString& nss,
                                         StringData ident,
                                         const RecordStore::Options& recordStoreOptions) {
-        auto status = createRecordStore(nss, ident, recordStoreOptions);
+        auto status = createRecordStore(provider, nss, ident, recordStoreOptions);
         if (status.isOK()) {
             return {ErrorCodes::DataModifiedByRepair, "Orphan recovery created a new record store"};
         }
@@ -375,6 +379,22 @@ public:
      */
     virtual void setJournalListener(JournalListener* jl) = 0;
 
+    /**
+     * See `StorageEngine::setLastMaterializedLsn`
+     */
+    virtual void setLastMaterializedLsn(uint64_t lsn) {}
+
+    /**
+     * Configures the specified checkpoint as the starting point for recovery.
+     */
+    virtual void setRecoveryCheckpointMetadata(StringData checkpointMetadata) {}
+
+    /**
+     * Configures the storage engine as the leader, allowing it to flush checkpoints to remote
+     * storage.
+     */
+    virtual void promoteToLeader() {}
+
     /**
      * See `StorageEngine::setStableTimestamp`
      */

src/mongo/db/storage/kv/kv_engine_test_harness.cpp

@@ -47,6 +47,7 @@
 #include "mongo/db/namespace_string.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/db/record_id.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/service_context_test_fixture.h"
 #include "mongo/db/storage/key_format.h"
 #include "mongo/db/storage/key_string/key_string.h"
@@ -116,8 +117,10 @@ protected:
         auto clientAndCtx = makeClientAndCtx("opCtx");
         auto opCtx = clientAndCtx.opCtx();
         KVEngine* engine = helper->getEngine();
+        auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
         ASSERT_OK(
-            engine->createRecordStore(NamespaceString::createNamespaceString_forTest("catalog"),
+            engine->createRecordStore(provider,
+                                      NamespaceString::createNamespaceString_forTest("catalog"),
                                       "collection-catalog",
                                       RecordStore::Options{}));
 
@@ -231,7 +234,8 @@ protected:
                                                 const RecordStore::Options& recordStoreOptions,
                                                 boost::optional<UUID> uuid) {
         auto opCtx = _makeOperationContext(engine);
-        ASSERT_OK(engine->createRecordStore(nss, ident, recordStoreOptions));
+        auto& provider = rss::ReplicatedStorageService::get(opCtx.get()).getPersistenceProvider();
+        ASSERT_OK(engine->createRecordStore(provider, nss, ident, recordStoreOptions));
         auto rs = engine->getRecordStore(opCtx.get(), nss, ident, recordStoreOptions, uuid);
         ASSERT(rs);
         return rs;
@@ -348,8 +352,14 @@ TEST_F(KVEngineTestHarness, SimpleSorted1) {
     {
         auto opCtx = _makeOperationContext(engine);
         auto& ru = *shard_role_details::getRecoveryUnit(opCtx.get());
-        ASSERT_OK(engine->createSortedDataInterface(
+        auto& provider = rss::ReplicatedStorageService::get(opCtx.get()).getPersistenceProvider();
-            ru, kNss, kUUID, kIdent, config, boost::none /* storageEngineIndexOptions */));
+        ASSERT_OK(engine->createSortedDataInterface(provider,
+                                                    ru,
+                                                    kNss,
+                                                    kUUID,
+                                                    kIdent,
+                                                    config,
+                                                    boost::none /* storageEngineIndexOptions */));
         sorted = engine->getSortedDataInterface(
             opCtx.get(), ru, kNss, kUUID, kIdent, config, kRecordStoreOptions.keyFormat);
         ASSERT(sorted);

src/mongo/db/storage/kv/kv_engine_timestamps_test.cpp

@@ -39,6 +39,7 @@
 #include "mongo/db/namespace_string.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/db/record_id.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/service_context.h"
 #include "mongo/db/service_context_test_fixture.h"
 #include "mongo/db/storage/kv/kv_engine.h"
@@ -237,7 +238,9 @@ public:
         const auto nss = NamespaceString::createNamespaceString_forTest("a.b");
         const auto ident = "collection-ident";
         RecordStore::Options options;
-        ASSERT_OK(engine->createRecordStore(nss, ident, options));
+        auto& provider =
+            rss::ReplicatedStorageService::get(getGlobalServiceContext()).getPersistenceProvider();
+        ASSERT_OK(engine->createRecordStore(provider, nss, ident, options));
         rs = engine->getRecordStore(op, nss, ident, options, UUID::gen());
         ASSERT(rs);
     }

src/mongo/db/storage/mdb_catalog.cpp

@@ -34,6 +34,7 @@
 #include "mongo/db/namespace_string.h"
 #include "mongo/db/operation_context.h"
 #include "mongo/db/record_id.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/storage/feature_document_util.h"
 #include "mongo/db/storage/kv/kv_engine.h"
 #include "mongo/db/storage/record_store.h"
@@ -242,7 +243,9 @@ StatusWith<std::unique_ptr<RecordStore>> MDBCatalog::createRecordStoreForEntry(
     const MDBCatalog::EntryIdentifier& entry,
     const boost::optional<UUID>& uuid,
     const RecordStore::Options& recordStoreOptions) {
-    Status status = _engine->createRecordStore(entry.nss, entry.ident, recordStoreOptions);
+    auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
+    Status status =
+        _engine->createRecordStore(provider, entry.nss, entry.ident, recordStoreOptions);
     if (!status.isOK()) {
         return status;
     }

src/mongo/db/storage/storage_engine.h

@@ -715,6 +715,26 @@ public:
      */
     virtual boost::optional<Timestamp> getLastStableRecoveryTimestamp() const = 0;
 
+    /**
+     * Sets the last materialized LSN, marking the highest phylog LSN
+     * that has been successfully written to the page server and should have no holes.
+     *
+     * TODO: Revisit how to handle cases where mongod speaks with a log server
+     * in a non-local zone due to failover.
+     */
+    virtual void setLastMaterializedLsn(uint64_t lsn) = 0;
+
+    /**
+     * Configures the specified checkpoint as the starting point for recovery.
+     */
+    virtual void setRecoveryCheckpointMetadata(StringData checkpointMetadata) = 0;
+
+    /**
+     * Configures the storage engine as the leader, allowing it to flush checkpoints to remote
+     * storage.
+     */
+    virtual void promoteToLeader() = 0;
+
     /**
      * Sets the highest timestamp at which the storage engine is allowed to take a checkpoint. This
      * timestamp must not decrease unless force=true is set, in which case we force the stable

src/mongo/db/storage/storage_engine_impl.cpp

@@ -38,6 +38,7 @@
 #include "mongo/db/local_catalog/catalog_raii.h"
 #include "mongo/db/local_catalog/shard_role_api/transaction_resources.h"
 #include "mongo/db/operation_context.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/storage/backup_cursor_hooks.h"
 #include "mongo/db/storage/deferred_drop_record_store.h"
 #include "mongo/db/storage/disk_space_monitor.h"
@@ -130,6 +131,15 @@ StorageEngineImpl::StorageEngineImpl(OperationContext* opCtx,
     invariant(prevRecoveryUnit->isNoop());
     shard_role_details::setRecoveryUnit(
         opCtx, _engine->newRecoveryUnit(), WriteUnitOfWork::RecoveryUnitState::kNotInUnitOfWork);
+
+    auto& rss = rss::ReplicatedStorageService::get(opCtx->getServiceContext());
+    if (rss.getPersistenceProvider().shouldDelayDataAccessDuringStartup()) {
+        LOGV2(10985326,
+              "Skip loading catalog on startup; it will be handled later when WT loads the "
+              "checkpoint");
+        return;
+    }
+
     // If we throw in this constructor, make sure to destroy the RecoveryUnit instance created above
     // before '_engine' is destroyed.
     ScopeGuard recoveryUnitResetGuard([&] {
@@ -181,8 +191,9 @@ void StorageEngineImpl::loadMDBCatalog(OperationContext* opCtx,
     if (!catalogExists) {
         WriteUnitOfWork uow(opCtx);
 
+        auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
         auto status = _engine->createRecordStore(
-            kCatalogInfoNamespace, ident::kMbdCatalog, catalogRecordStoreOpts);
+            provider, kCatalogInfoNamespace, ident::kMbdCatalog, catalogRecordStoreOpts);
 
         // BadValue is usually caused by invalid configuration string.
         // We still fassert() but without a stack trace.
@@ -422,8 +433,9 @@ Status StorageEngineImpl::_recoverOrphanedCollection(OperationContext* opCtx,
     WriteUnitOfWork wuow(opCtx);
     const auto recordStoreOptions =
         _catalog->getParsedRecordStoreOptions(opCtx, catalogId, collectionName);
-    Status status =
+    auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
-        _engine->recoverOrphanedIdent(collectionName, collectionIdent, recordStoreOptions);
+    Status status = _engine->recoverOrphanedIdent(
+        provider, collectionName, collectionIdent, recordStoreOptions);
 
     bool dataModified = status.code() == ErrorCodes::DataModifiedByRepair;
     if (!status.isOK() && !dataModified) {
@@ -706,6 +718,18 @@ void StorageEngineImpl::setJournalListener(JournalListener* jl) {
     _engine->setJournalListener(jl);
 }
 
+void StorageEngineImpl::setLastMaterializedLsn(uint64_t lsn) {
+    _engine->setLastMaterializedLsn(lsn);
+}
+
+void StorageEngineImpl::setRecoveryCheckpointMetadata(StringData checkpointMetadata) {
+    _engine->setRecoveryCheckpointMetadata(checkpointMetadata);
+}
+
+void StorageEngineImpl::promoteToLeader() {
+    _engine->promoteToLeader();
+}
+
 void StorageEngineImpl::setStableTimestamp(Timestamp stableTimestamp, bool force) {
     _engine->setStableTimestamp(stableTimestamp, force);
 }

src/mongo/db/storage/storage_engine_impl.h

@@ -144,6 +144,12 @@ public:
 
     void cleanShutdown(ServiceContext* svcCtx, bool memLeakAllowed) override;
 
+    void setLastMaterializedLsn(uint64_t lsn) override;
+
+    void setRecoveryCheckpointMetadata(StringData checkpointMetadata) override;
+
+    void promoteToLeader() override;
+
     void setStableTimestamp(Timestamp stableTimestamp, bool force = false) override;
 
     Timestamp getStableTimestamp() const override;

src/mongo/db/storage/storage_engine_mock.h

@@ -129,7 +129,14 @@ public:
     boost::optional<Timestamp> getLastStableRecoveryTimestamp() const final {
         MONGO_UNREACHABLE;
     }
-    void setStableTimestamp(Timestamp stableTimestamp, bool force = false) final {}
+
+    void setLastMaterializedLsn(uint64_t lsn) final {}
+
+    void setRecoveryCheckpointMetadata(StringData checkpointMetadata) final {}
+
+    void promoteToLeader() final {}
+
+    void setStableTimestamp(Timestamp stableTimestamp, bool force = false) override {}
     Timestamp getStableTimestamp() const override {
         return Timestamp();
     }

src/mongo/db/storage/storage_engine_test_fixture.h

@@ -36,6 +36,7 @@
 #include "mongo/db/local_catalog/durable_catalog.h"
 #include "mongo/db/local_catalog/shard_role_api/transaction_resources.h"
 #include "mongo/db/repl/storage_interface_impl.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/service_context_d_test_fixture.h"
 #include "mongo/db/storage/kv/kv_engine.h"
 #include "mongo/db/storage/mdb_catalog.h"
@@ -117,8 +118,9 @@ public:
      */
     Status createCollTable(OperationContext* opCtx, NamespaceString collName) {
         const std::string identName = _storageEngine->generateNewCollectionIdent(collName.dbName());
+        auto& provider = rss::ReplicatedStorageService::get(opCtx).getPersistenceProvider();
         return _storageEngine->getEngine()->createRecordStore(
-            collName, identName, RecordStore::Options{});
+            provider, collName, identName, RecordStore::Options{});
     }
 
     Status dropIndexTable(OperationContext* opCtx, NamespaceString nss, StringData indexName) {

src/mongo/db/storage/storage_options.cpp

@@ -45,10 +45,14 @@
 namespace mongo {
 
 StorageGlobalParams::StorageGlobalParams() {
-    reset();
+    _reset();
 }
 
-void StorageGlobalParams::reset() {
+void StorageGlobalParams::reset_forTest() {
+    _reset();
+}
+
+void StorageGlobalParams::_reset() {
     engine = "wiredTiger";
     engineSetByUser = false;
     dbpath = kDefaultDbPath;
@@ -60,7 +64,7 @@ void StorageGlobalParams::reset() {
 
     noTableScan.store(false);
     directoryperdb = false;
-    syncdelay.store(60.0);
+    syncdelay.store(-1.0);
     queryableBackupMode = false;
     groupCollections = false;
     oplogMinRetentionHours.store(0.0);

src/mongo/db/storage/storage_options.h

@@ -44,7 +44,7 @@ namespace mongo {
 
 struct StorageGlobalParams {
     StorageGlobalParams();
-    void reset();
+    void reset_forTest();
 
     // Returns the directory path used by the spill storage engine to store spilled data.
     std::string getSpillDbPath() const;
@@ -109,13 +109,14 @@ struct StorageGlobalParams {
 
     // --syncdelay
     // Delay in seconds between triggering the next checkpoint after the completion of the previous
-    // one. A value of 0 indicates that checkpointing will be skipped.
+    // one. A value of 0 indicates that checkpointing will be skipped. A value <0
+    // will result in using the default value for the configured persistence provider.
     // Do not set this value on production systems.
     // In almost every situation, you should use the default setting.
     // This parameter is both a server parameter and a configuration parameter, and to resolve
-    // conflicts between the two the default must be set here.
+    // conflicts between the two, a default sentinel (<0) must be set here.
     static constexpr double kMaxSyncdelaySecs = 60 * 60;  // 1hr
-    AtomicWord<double> syncdelay{60.0};                   // seconds between checkpoints
+    AtomicWord<double> syncdelay{-1.0};                   // seconds between checkpoints
 
     // --queryableBackupMode
     // Prevents user-originating operations from performing writes to the server. Internally
@@ -139,6 +140,9 @@ struct StorageGlobalParams {
 
     // Test-only option. Disables table logging.
     bool forceDisableTableLogging = false;
+
+private:
+    void _reset();
 };
 
 extern StorageGlobalParams storageGlobalParams;

src/mongo/db/storage/wiredtiger/BUILD.bazel

@@ -112,6 +112,7 @@ mongo_cc_library(
         "//src/mongo/db:server_base",
         "//src/mongo/db:server_feature_flags",
         "//src/mongo/db:service_context",
+        "//src/mongo/db/rss:replicated_storage_service",
         "//src/mongo/db/storage:container",
         "//src/mongo/db/storage:exceptions",
         "//src/mongo/db/storage:execution_context",
@@ -173,6 +174,7 @@ mongo_cc_library(
     deps = [
         ":storage_wiredtiger_core",
         "//src/mongo/db:service_context_test_fixture",
+        "//src/mongo/db/rss:persistence_provider_impl",
         "//src/mongo/db/storage:record_store_test_harness",
         "//src/mongo/util:clock_source_mock",
     ],
@@ -228,6 +230,7 @@ mongo_cc_unit_test(
         "//src/mongo/db/storage:storage_options",
         "//src/mongo/db/storage/kv:kv_engine_test_harness",
         "//src/mongo/idl:server_parameter_test_controller",
+        "//src/mongo/idl:server_parameter_test_util",
         "//src/mongo/util:clock_source_mock",
     ],
 )

src/mongo/db/storage/wiredtiger/spill_wiredtiger_kv_engine.cpp

@@ -69,7 +69,7 @@ SpillWiredTigerKVEngine::SpillWiredTigerKVEngine(const std::string& canonicalNam
     }
 
     std::string config =
-        generateWTOpenConfigString(_wtConfig, wtExtensions.getOpenExtensionsConfig());
+        generateWTOpenConfigString(_wtConfig, wtExtensions.getOpenExtensionsConfig(), "");
     LOGV2(10158000, "Opening spill WiredTiger", "config"_attr = config);
 
     auto startTime = Date_t::now();

src/mongo/db/storage/wiredtiger/spill_wiredtiger_kv_engine.h

@@ -94,7 +94,8 @@ public:
         MONGO_UNREACHABLE;
     }
 
-    Status createRecordStore(const NamespaceString& nss,
+    Status createRecordStore(const rss::PersistenceProvider&,
+                             const NamespaceString& nss,
                              StringData ident,
                              const RecordStore::Options& options) override {
         MONGO_UNREACHABLE;
@@ -126,6 +127,7 @@ public:
     }
 
     Status createSortedDataInterface(
+        const rss::PersistenceProvider&,
         RecoveryUnit&,
         const NamespaceString& nss,
         const UUID& uuid,

src/mongo/db/storage/wiredtiger/spill_wiredtiger_kv_engine_test.cpp

@@ -32,10 +32,12 @@
 
 #include "mongo/base/init.h"  // IWYU pragma: keep
 #include "mongo/base/string_data.h"
+#include "mongo/db/rss/replicated_storage_service.h"
 #include "mongo/db/service_context.h"
 #include "mongo/db/service_context_test_fixture.h"
 #include "mongo/db/storage/wiredtiger/wiredtiger_extensions.h"
 #include "mongo/db/storage/wiredtiger/wiredtiger_global_options_gen.h"
+#include "mongo/idl/server_parameter_test_util.h"
 #include "mongo/unittest/temp_dir.h"
 #include "mongo/unittest/unittest.h"
 #include "mongo/util/clock_source_mock.h"

src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.cpp

@@ -34,6 +34,7 @@
 #include "mongo/db/service_context.h"
 #include "mongo/util/assert_util.h"
 #include "mongo/util/decorable.h"
+#include "mongo/util/str.h"
 
 #include <memory>
 #include <utility>
@@ -49,8 +50,33 @@ ServiceContext::ConstructorActionRegisterer setWiredTigerCustomizationHooks{
 
 const auto getCustomizationHooks =
     ServiceContext::declareDecoration<std::unique_ptr<WiredTigerCustomizationHooks>>();
+
+const auto getWiredTigerCustomizationHooksRegistry =
+    ServiceContext::declareDecoration<WiredTigerCustomizationHooksRegistry>();
+
 }  // namespace
 
+
+WiredTigerCustomizationHooksRegistry& WiredTigerCustomizationHooksRegistry::get(
+    ServiceContext* service) {
+    return getWiredTigerCustomizationHooksRegistry(service);
+}
+
+
+void WiredTigerCustomizationHooksRegistry::addHook(
+    std::unique_ptr<WiredTigerCustomizationHooks> custHook) {
+    invariant(custHook);
+    _hooks.push_back(std::move(custHook));
+}
+
+std::string WiredTigerCustomizationHooksRegistry::getTableCreateConfig(StringData tableName) const {
+    str::stream config;
+    for (const auto& h : _hooks) {
+        config << h->getTableCreateConfig(tableName);
+    }
+    return config;
+}
+
 void WiredTigerCustomizationHooks::set(ServiceContext* service,
                                        std::unique_ptr<WiredTigerCustomizationHooks> customHooks) {
     auto& hooks = getCustomizationHooks(service);

src/mongo/db/storage/wiredtiger/wiredtiger_customization_hooks.h

@@ -31,6 +31,7 @@
 
 #include <memory>
 #include <string>
+#include <vector>
 
 namespace mongo {
 class StringData;
@@ -58,4 +59,27 @@ public:
     virtual std::string getTableCreateConfig(StringData tableName);
 };
 
+/**
+ * Registry to store multiple WiredTiger customization hooks.
+ */
+class WiredTigerCustomizationHooksRegistry {
+public:
+    static WiredTigerCustomizationHooksRegistry& get(ServiceContext* serviceContext);
+
+    /**
+     * Adds a WiredTiger customization hook to the registry. Multiple hooks can be
+     * added, and their configurations will be combined.
+     */
+    void addHook(std::unique_ptr<WiredTigerCustomizationHooks> custHook);
+
+    /**
+     * Gets a combined configuration string from all hooks in the registry for
+     * the provided table name during the `WT_SESSION::create` call.
+     */
+    std::string getTableCreateConfig(StringData tableName) const;
+
+private:
+    std::vector<std::unique_ptr<WiredTigerCustomizationHooks>> _hooks;
+};
+
 }  // namespace mongo

src/mongo/db/storage/wiredtiger/wiredtiger_global_options.h

@@ -69,6 +69,7 @@ public:
     std::string liveRestoreSource;
     int liveRestoreThreads;
     double liveRestoreReadSizeMB;
+    int flattenLeafPageDelta;
 
     std::string collectionBlockCompressor;
     bool useIndexPrefixCompression;

src/mongo/db/storage/wiredtiger/wiredtiger_global_options.idl

@@ -126,6 +126,18 @@ server_parameters:
             lte: 100
         redact: false
 
+    wiredTigerFlattenLeafPageDelta:
+        description: >-
+            WiredTiger page read rewrites the leaf pages with deltas to a new disk image if 
+            successful. We will use a ternary where 0=Disabled, 1=Enabled if disaggregatedStorageEnabled is true, 2=Enabled unconditionally.
+        set_at: startup
+        cpp_varname: "wiredTigerGlobalOptions.flattenLeafPageDelta"
+        default: 1
+        validator:
+            gte: 0
+            lte: 2
+        redact: false
+
     wiredTigerEvictionDirtyTargetGB:
         description: >-
             Absolute dirty cache eviction target. Once eviction begins,